The association of garlic intake and cardiovascular risk factors: A systematic review and meta-analysis

Abstract Garlic is a common cooking ingredient and used in traditional medicine in Asian countries. There is a growing attention on garlic due to its preventive characteristics in cardiovascular diseases (CVDs). Many studies have reviewed the association between garlic intake and CVDs; however, no consistent conclusions have been drawn. New clinical trials have also been conducted and could contribute to more solid statements. In order to systematically review the reliability of previous studies regarding the implication of garlic in the management of CVDs, we performed in-depth meta-analysis using the most up-to-date randomized clinical trials (RCTs) data with more systematic controls. According to the 22 studies included, the effects of garlic intake on lowering total cholesterol (TC) and low-density lipoprotein (LDL) are more noticeable with lower dosage and longer duration, especially in patients with CVDs. In addition, subgroup analysis indicated that appropriate diet intervention could be an important control factor that should be taken into consideration in any future study designs.


Introduction
Cardiovascular diseases (CVDs) are the leading cause of death worldwide (Virani et al. 2020).The term CVD refers to a group of disorders of the heart and blood vessels.It commonly includes diseases such as coronary heart disease, heart failure, cardiomyopathy, congenital heart disease, peripheral vascular disease and stroke.Many of these conditions are life-threatening.Major risk factors of CVDs are obesity, hypertension, dyslipidemia, impaired glucose tolerance, raised levels of inflammatory markers, such as C-reactive protein (CRP), cytokines, tumor necrosis factor alpha (TNF-α), interleukins 6 and 10 (IL-6 and IL-10), changes in cell adhesion molecules, prothrombotic and fibrinolytic changes, an increase in oxidative stress and endothelial dysfunction (Juturu and Gormley 2013).
Previous in-vivo and in-vitro studies have reported that garlic has potential antilipidemic, antihypertensive, antiglycemic, antithrombotic and antiatherogenic properties (Ansary et al. 2020).Due to this, garlic has been manufactured into various herbal health products and has been used to improve heart health and provide immunity support.The antilipidemic role of garlic intake has been controversial.While studies from Berthold et al. and Isaacsohn et al. show negative findings (Berthold, Sudhop, and von Bergmann 1998;Isaacsohn et al. 1998), several recent meta-analyses and reviews have suggested that garlic intake is associated with a dramatic drop of TC, systolic blood pressure (SBP) and diastolic blood pressure (DBP) with a potential increase in LDL (Ribeiro et al. 2021).However, these latter studies were either with obsolete data, partially reviewed, have bias on RCT selection, or have not performed full-validation tests.
Therefore, to have a better insight into if garlic intake reduces cardiovascular risks, we performed a thorough meta-analysis on the most recent RCTs.We examined the effects of garlic intakes on patients with CVDs or individuals with CVD-associated risk factors, including hypercholesterolemia, hypertension and obesity.Key measurements covered four major aspects: (1) the lipid-associated markers, such as TC, triglyceride (TG), LDL, high density lipoprotein (HDL), homocysteine, apolipoprotein A-1 (ApoA1), apolipoprotein B (ApoB) and lipoprotein a (Lp(a)); (2) blood pressure associated markers, such as SBP and DBP; (3) markers indicating liver functions such as alanine aminotransferase (ALT), aspartate aminotransferase (AST), alkaline phosphatase (ALP) and (4) blood sugar/glucose, and inflammatory markers, such as CRP, TNF-α, IL-6 and fibrinogen.
Although diabetes melllitus and kidney dysfunctions share risk factors with CVDs to a certain extent, they may associate with other non-CVD risk factors and hence were excluded in our study.Furthermore, we performed additional analyses to validate our results, including sensitivity tests, inter-and intra-group heterogeneity tests, publication bias examination with three methods and further quality assessments between sub-groups.Hence, the findings and conclusions in this study were drawn in accordance to the best meta-analysis practice guidelines.The results of our detailed and robust analyses provide further evidence that garlic intake would lead to reduced TC levels.Garlic is also likely to also reduce the LDL and SBP levels, but more RCTs would need to be conducted for a more reliable conclusion.
For lipid metabolisms, more research weight was put on the anti-inflammatory properties of garlic (Arreola et al. 2015;Moosavian et al. 2020).We also found a decrease in IL-6 and TNF-α.However, due to a limited number of studies available, further validation would also be required.Interestingly, our sub-group analysis of with and without diet intervention showed that diet intervention is an essential prerequisite for garlic intake studies to have significant effects on TC and LDL levels.Hence, this can serve as a guide and a basis for further RCT researches.

Search strategy
The study was performed following the Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA checklist).Protocol of this study was registered in International Prospective Register of Systematic Reviews (PROSPERO) with registration number CRD42021272199.Literature databases, including PubMed, Web of Knowledge, EMBASE, Wanfang and CNKI were searched through studies that were published from 1990 to date (2021).Key search terms were garlic and Allium sativum combined with cardiovascular-related factors, such as cardiovascular diseases, cholesterol, blood glucose, LDL, HDL, blood pressure, DBP, SBP, inflammation, hypertension, hypercholesterolemia, hyperlipidemia, atherosclerosis and blood circulation.Since this review was focused on CVDs and their corresponding risk factors, other diseases that share risk factors with CVDs, such as diabetes mellitus and renal diseases, were not considered.Each publication was independently reviewed by two reviewers.

Inclusion criteria
1. Participants diagnosed as hypercholesterolemia, hypertension, obesity, CVDs or populations that have high risk for CVDs (such as 10-year Framingham risk > 10%).2. Study participants to be at least 18 years old.3. RCTs involved garlic products for intervention, including raw garlic, dried garlic powder, alliin, allicin, AGE and garlic oil.

Exclusion criteria
Studies were excluded if: (1) patients suffered from other known diseases such as cancer, diabetes, kidney dysfunctions; (2) studies that do not report any efficacy measurements; (3) any non-human and animal works or any cross-over designs or studies without a control/placebo group; (4) studies with biased starting population, such as the ones on postmenopausal women, pregnant women and trained men; (5) studies with missing information that cannot be converted or imputed; (6) studies that are reviews, conference abstracts, posters, unpublished data, commentaries and theses.

Data extraction
A standardized data extraction form was established, and two investigators independently extracted the data using the same form.

Quality assessment
The data quality was assessed independently by two investigators, according to the quality assessment tool in RevMan5.

Statistical analysis
For detailed data conversion and imputation approaches, please refer to the supplementary information.To conduct a fair analysis, only one mean from each of the garlic-treated group and the control group was extracted.If multiple intermediate results were available, the one with the highest dose and the most prolonged treatment period was selected for the following analysis.
The between-group mean difference (MD between ) was calculated as:

MD
x x between trt c tl where x trt is the mean of the garlic-treated group and x ctl is the mean of the control/placebo group.All values were measured at the end of the treatment period.
The between-group standardized mean difference ( SMD between ) the expresses the difference between the garlic-treated group and the control group in units of standard deviation was calculated as: where S pooled is the pooled standard deviation of both groups and was calculated as: where n trt and s trt represents the sample size and the standard deviation in the garlic-treated group, respectively and n ctl and s ctl represents the sample size in the control group, respectively.
According to the conventions by Cohen (1998), SMD between was interpreted as: • SMD ∼ 0.2: small effect • SMD ∼ 0.5: moderate effect • SMD ∼ 0.8: large effect The main analysis was done using the meta R package.Both fixed-effect model that assumes that all studies share the same true effect size and random-effect model that provides more variance were applied to the measured terms, including cardiovascular markers: ApoA1, ApoB, HDL, LDL, TC, TG, Lp(a) and homocysteine; inflammatory markers: CRP, fibrinogen, IL-6 and TNF-α; blood pressure indicators: DBP and SBP; and markers of liver function: ALP, ALT, AST and blood sugar.For the random-effect models, the between-study variance was estimated by the restricted maximum likelihood (REML, τ 2 ) method and the method by Hartung and Knapp was used to adjust test statistics and confidence intervals.The Higgins & Thompson's I 2 that quantifies between-study heterogeneity was calculated for both models based on the Cochran's Q.The fixed-effect models results were shown on the forest plots with the RevMan5 layout.Between-study heterogeneity was considered as low if I 2 < 30%, moderate if I 2 is between 30-50% and high if I 2 > 75%.
Publication bias was represented using funnel plots.Significant thresholds including p < 0.1, 0.05 and 0.01 were provided as contours and effect sizes that fell into p < 0.05 and 0.01 regions were considered significant (Peters et al. 2008).The funnel plot asymmetry was tested using three different methods, including the Egger's regression test, the Begg's test and the Macaskill test.Since the p values of LDL and TG were significant that show potential publication bias, Duval & Tweedie Trim and Fill method was used to perform bias-corrected estimation of the true effect sizes (Duval and Tweedie 2000).In addition, the outliers that overly contributed to the heterogeneity in the analysis were predicted using the find.outliersfunction in the dmetar package and represented by the baujat plots (Baujat et al. 2002).
For subgroup effect analysis, mixed-effect models were used to assess the true difference between subgroups.Subgroups were defined for dose of AGE intake, duration of the study, type of diseases and location of study.This assumed random effects within subgroups and contained also fixed effects since subgroups were assumed to be fixed.In addition, subgroup analyses were conducted without a common estimate of τ 2 across subgroups.To define subgroups of dose and duration, density plots were used for visualization using the density function in R. The effect sizes, 95% confidence intervals (CI), p values, I 2 and their corresponding 95% CI in each subgroup as well as the p value of the test for subgroup differences were reported in the relevant tables.In addition, to evaluate the quality of the study, a sensitivity analysis was carried out for the assessment of publication year, if the study was placebo-controlled, if the study was conducted as double-blinded, if the quality of the study was high (according to the overall score of the quality assessment), if a diet intervention was conducted and if the garlic product was branded.A p value of < 0.05 was considered significant.
To estimate the trend over different doses and duration, all available data were used to fit the linear model.Therefore, studies that were conducted with multi-time points and multi-dosage would have all data points considered rather than what was used in the effect size estimation, in which only the data point with the maximum dose and time duration was selected.To minimize the impacts of results with high heterogeneity, only results within 1%-99% quantile range were kept.The scatter plot was done using the ggplot2 package in R. The linear fit was done using the geom_ smooth function and the pearson correlation was calculated using the stat_cor function.

Characteristics of included studies
A total of 3336 studies were retrieved in the initial database searching.After removing duplicated studies, 1549 records remained.After further review of the abstracts, 1291 records were excluded if they were:retrospective studies, non-human RCTs, literature reviews or studies that were not relevant to garlic.258 records were eligible for further consideration and only of these, 69 included RCTs.After a full-text review, a further 47 studies were excluded due to limitations indicated in Table S1, based on pre-selected exclusion criteria.In total 22 studies were used for meta-analysis.A schematic view of the study selection criteria is shown in Figure 1.
From the 22 RCTs, a total of 1333 participants were enrolled; 752 received garlic interventions while 581 were in the control group receiving placebos.All patients included in these studies were assessed via ultrasound examination and/or biochemical test and were either diagnosed with CVDs or at high risks of developing CVDs.All studies showed comparable baseline data.The characteristics and baseline data of our included studies were listed in Table 1.The risk of bias of recruited studies was evaluated in accordance to the scoring system developed by the Cochrane Collaboration (Figures 2 and 3).Applying random grouping using a computed random number or referring to a random number table indicated that the majority of studies showed low risk of selection bias in terms of random sequence generation.The blinding of outcome assessment was also performed in most pooled studies.Missing information in allocation concealment and blinding of participants and personnel may lead to unclear/high risk of selection bias, performance bias and detection bias.Overall, the quality of the pooled studies was satisfactory i.e. presented low risk of bias despite high heterogeneity between the independent trails.Data in all 22 RCTs was therefore used in further meta-analysis.

Blood pressure
The effects of garlic on blood pressure were also evaluated (Figure 7).Hypertension is commonly defined with two measurements, with SBP > = 140 mmHg and DBP > = 90 mmHg.SBP is a more frequent CVD risk factor than DBP (Banegas et al. 2002).A total of 7 studies enrolling 416 patients (212 garlic group and 204 placebo group) reported whether garlic impacted DBP and SBP.After meta-analysis, results indicated that garlic intervention significantly reduced the level of SBP (SMD = −0.27,95% CI = −0.46-0.07,I 2 = 66%, Figure 7A) while having no obvious effects on the change of DBP (SMD = −0.02,95% CI = −0.22-0.17,I 2 = 75%, Figure 7B).However, Aslani et al. (2016) contributed markly to the high heterogeneity among the studies and if removed, the SBP level turned to be non-significant.

Liver functions
Nonalcoholic fatty liver disease (NAFLD) is consistently associated with features of metabolic syndrome, which is a condition with a high risk of cardiovascular events  (Godoy-Matos, Silva Junior, and Valerio 2020).Particularly, several commonly used markers of liver dysfunction, such as ALT, AST and ALP, have also been implicated with the risk of CVDs (Farrell, Haczeyni, and Chitturi 2018; Godoy-Matos, Silva Junior, and Valerio 2020).Three of our included studies reported the effects of garlic on liver functions in populations with hypercholesterolemia (Sobenin et al. 2008), hypertension (Ried, Travica, and Sali 2016) or known coronary artery disease (Budoff et al. 2004).From our meta-analysis, there was no significant improvements on ALT (SMD = −0.20,95% CI = −0.52-0.13,I 2 = 0%, 149 patients, Figure 8A), and AST (SMD = −0.12,95% CI = −0.44-0.21,I 2 = 4%, 149 patients, Figure 8B) with garlic intake; the level of ALP was slightly increased (SMD = −0.41,95% CI = −0.14-0.33,I 2 = 0%, 280 patients, Figure 8C).In addition to liver enzymes, the blood sugar levels regulated by the liver is another parameter associated with increased risk of CVDs (Patel et al. 2016;Tzoulaki et al. 2016).Three pooled studies have reported the effect of garlic on blood sugar content in similar populations mentioned above.Meta-analysis results showed no significant change in the   level of blood sugar (SMD = 0.21, 95% CI = −0.1-0.51,I 2 = 0%, 166 patients, Figure 8D).However, due to the limited number of studies and patients involved, a conclusive result to whether garlic has effects on regulating liver function enzymes and blood sugar levels are difficult to ascertain.

Inflammation
Increasing evidence indicate that systemic inflammation serves as a nontraditional risk factor for CVDs (Seo et al. 2014).CVDs are associated with elevated markers of systemic inflammation, including C-reactive protein (CRP) and members of the coagulation cascades.Elevated levels of CRP and fibrinogen, production of tumor necrosis factor (TNF)-α, and IL-6 are associated with increased risk of coronary heart disease.Four studies with 206 participants (106 in the garlic intervention group and 100 in the control group) presented the effect of garlic on TNF-α level in serum.The meta-analysis of these investigations showed a significant reduction of TNF-α in garlic intervention group compared with placebo group (SMD = −0.48,95% CI = −0.76-0.2,I 2 = 14%, 206 patients, p value for overall effect <0.01, Figure 9A).Significant reduction on serum level of IL-6 was also found (SMD = −0.77,95% CI = −1.16-0.38,I 2 = 90%, 120 patients, p value for overall effect <0.01, Figure 9B) despite high heterogeneity between the studies.For fibrinogen (SMD = −0.16,95% CI = −0.47-0.14,I 2 = 63%, 169 patients, p value for overall effect = 0.29, Figure 9C) and CRP (SMD = 0.06, 95% CI = −0.17-0.28,I 2 = 9%, 310 patients, p value for overall effect = 0.61, Figure 9D) levels, no significant changes were observed.Due to the limited number of studies and high heterogeneity, more high quality RCTs with large sample sizes would be required to form a definitive conclusion.

Meta regression
We conducted meta-regression to further analyze several modulators such as garlic dosage, intervention duration, baseline value of the garlic groups, and subject number.Considering the data availability, only TC, TG, HDL, LDL, SBP and DBP that had more than 7 studies were included in these analyses.Our results indicated that the effects of garlic on most of these variables were not moderated by dosage, duration, baseline value and subject number.Notably, intervention duration was inversely correlated with the effect size of TC (R = −0.42,p = 0.022) (Figure 10A, Figures S3A and S5A).Terms TG and HDL did not show any significant difference in effect sizes between the garlic and control groups.However, the effect sizes of TG could potentially be affected by the baseline values of garlic groups (R = 0.48, p = 0.00075) (Figure 10C, Figures S3C and  S5C) and the number of participants could have an impact on the LDL effect sizes since they have a positive correlation of 0.5 with p = 0.0046 (Figure 11A, Figures S4A and  S6A).Although dosage was negatively correlated with the effect size of HDL (R = −0.41,p = 0.0041), this was likely to be mainly contributed by Gardner's studies since they performed the assessments with multiple doses (Figure 10B, Figures S3B and S5B).

Sensitivity/subgroup analyses
To evaluate whether the results were driven by a single study, we performed a sensitivity test by removing one study at a time sequentially.The sensitivity results showed that omission of any study did not change our main findings of TC, LDL, and SBP, indicating that these results were statistically stable (Figure 12A).We have performed sub-group analysis for the major terms, including TC, TG, HDL, LDL, SBP and DBP.Subgroups of dosage, duration and publication year were defined based on the density plots (Figure S2).The obese group was not reported due to the limited number of studies.Asia was separated since garlic is a major flavoring herb used in Asian cuisine.It was worth mentioning that the regression analysis would put more weight on the studies with multiple durations  and doses.This potential bias should be minimized in the main and subgroup analysis since we have restricted the data selected to the one with the maximum dosage and duration.It was found that, for the terms TC and LDL that show a degree of significance in the main and regression analysis, such significance was contributed by lower dose (<1100 mg AGE equivalent), longer duration (> =90 days) and in patients with CVDs.The location did not seem to be a major contributor to the final results (Table 2).

Restriction sensitivity
Additional sensitivity tests were conducted on the more recent publications (Publication year > = 2010), studies with randomized sampling which we term "high-quality groups", double-blinded studies and RCTs with diet intervention or used "branded" garlic products (Table 3).For the terms that showed an effect, such as TC and LDL, although the branded and high-quality groups still maintained the significance, the more recent and double-blinded studies disagreed with the initial results.Therefore, these analyses will require further thorough examination and interpretation Diet intervention could be an important impact factor as the significance was striking (p = 0.03) with extremely low heterogeneity I 2 percentage (I 2 = 0).

Publication bias and data heterogeneity tests
Funnel plots were used to evaluate the publication bias of our selected studies.The asymmetries of the funnel plots were validated using 3 different approaches: the Egger's regression test, Begg's test and the Macaskill's test.We assessed the CVD risk factors in Figure 13, including ALT, AST, ApoA1, ApoB, Blood sugar, CRP, DBP, fibrinogen, HDL, homocysteine, IL-6, LDL, SBP, TC, TG, and TNF-α.The results showed high publication bias potential in terms of LDL and TG.The contribution of overall heterogeneity and influence on pooled results was also calculated for the terms measured by at least three studies, including TC, TG, LDL and HDL.Effect sizes and p values were re-estimated after removing outlier studies and compared with the original ones.Forest plots were included for TG and LDL with high publication bias potential.It was found that publication bias and data heterogeneity have large impacts on the final result and need to be carefully considered before drawing any informative conclusion.

Discussion
Garlic is one of the most widely used cooking herbs and has been manufactured into popular healthcare products worldwide.There are various forms of garlic products in the market, including garlic powder, garlic oil and AGE.
In-vitro studies have reported the antilipidemic, antihypertensive, antiglycemic, antithrombotic and antiatherogenic properties of garlic (Ansary et al. 2020;Li et al. Figure 12. (a) sensitivity test by removing one study at a time and re-assess if the effect sizes are significant using the fixed-effect models for the major terms in four sectors that are associated with Cvds, including liver function (lF), blood lipid levels (Bl), inflammation status (iM) and blood pressure (BP).(B) Publication bias assessments.tests were done with (1) all studies; (2) after trimming and filtering (tf ) using the trim and fill method; (3) after removing the outliers (rm) that are outside 95% confidence interval of the pooled effect; (4) after both tf and rm.P value of 0.05 is indicated in the plot using the dashed orange line.
2013; Superko and Kruss 2000;Zeng et al. 2013).However, whether garlic is capable of reducing the blood lipid level is still controversial.Although Berthold, Sudhop, and von Bergmann (1998) and Isaacsohn et al. (1998) argued with garlic's antilipidemic claim back-to-back, several studies have since provided conflicting evidence (Ribeiro et al. 2021;Schwingshackl, Missbach, and Hoffmann 2016).Furthermore, the main advantages of garlic intake are still unclear.As a healthcare product advertised in improving cardiovascular health and immunity function, would existing RCTs support the preventive role of garlic in CVDs and in which aspects do garlic products fulfill such statements in terms of CVD-associated risk factors?To answer these questions and have a better view of the mechanisms under garlic intake in the cardiovascular field, we reviewed and thoroughly analyzed the most recent 22 RCTs and investigated if garlic intake could reduce the CVD risk factors.In total, our meta-analysis included 1333 participants recruited from 15 different regions around the world.All studies were carefully selected so that only patients with CVDs or individuals with key CVD-associated risk factors were considered, such as hypercholesterolemia, hypertension and obesity.Diabetes mellitus and renal diseases that share risk factors with CVDs were excluded since such disease conditions could be contributed by other non-CVD associated factors.In this meta-analysis study, we found that garlic intervention significantly reduced blood TC level while no obvious changes among TG levels were found.Moreover, a significant reduction of TNF-α and IL-6 in the garlic intervention group compared with the placebo group was observed, indicating the protective effect of garlic on alleviating inflammation.Our sensitivity test results showed that omission of any study did not change our main findings of TC, LDL, and SBP, indicating that these results were statistically stable.Additional sensitivity tests indicated that diet intervention could be an important impact factor for the effects of garlic as the significance was striking with extremely low heterogeneity, suggesting that garlic intake combined with having a low-cholesterol or low-fat diet is highly recommended.Our meta-regression analysis showed that the effects of garlic on most of the major measured outcomes were not moderated by the garlic dosage, intervention duration, baseline value and subject number.Notably, we found that the effect sizes of TG could potentially be affected by the baseline values of garlic groups, and the number of participants could have an impact on the HDL effect sizes.To have a more comprehensive comparison, we have performed a sub-group analysis for the major terms, including TC, TG, HDL, LDL, SBP and DBP.In terms of TC and LDL, we noted that the significant results were contributed by lower dose (<1100 mg AGE equivalent), longer duration (>90 days) in patients with CVDs.There were no obvious changes among the homocysteine levels, Lp(a), ApoA1, ApoB, ALT, AST, blood sugar, fibrinogen and CRP.However, due to the limited number of studies and patients involved, a conclusive result to whether garlic showed effects on regulating these factors is inconclusive.
It is worth mentioning that some data will require further scrutiny and investigation.For example, LDL level showed a reduction in the garlic groups compared to the control.However, considering the trace of publication bias, the impact was not significant after either removing the heterogeneous outliers or after data trimming and filling.Conversely for HDL level, garlic groups showed a slight increase but without statistical significance; however, after the trimming and filling procedure, the effect of garlic on HDL level was significant.Additionally, results indicated that garlic intervention significantly reduced the level of SBP.However, the study performed by Aslani et al. contributed the most to the high heterogeneity among the studies (Aslani et al. 2016) and the effect of SBP level became non-significant if this study (Aslani et al. 2016) was removed.
This meta-analysis would be the first to include all the available measurements reported in the high-quality RCTs associated with CVDs and its risk factors with garlic supplements as the treatment product to date.In contrast to other available meta-analyses and systematic reviews (Emamat et al. 2020;Schwingshackl, Missbach, and Hoffmann 2016), our study provides thorough analysis and more systematic controls.We examine publication bias, between-study and within-study heterogeneity with different methods and have validated our findings by removing the outliers that result in high heterogeneity.In addition, sensitivity tests were performed by testing the impact of removing one study each time, ensuring that one particular study did not significantly influence our major findings.Further quality assessment indicated that overall changes in TC and LDL should be more carefully interpreted due to high heterogeneity.The effects of garlic intake on TC and LDL are more noticeable with a lower dose, longer duration, especially in patients with CVDs.To be noted, diet intervention could be an important control factor since subgroup analysis showed a more dramatic increase of effect size and a huge drop of heterogeneity with consideration of healthy functional food and dietary supplements for conventional diet.Although all the data have been extracted and analyzed with caution, several limitations would need to be acknowledged.Firstly, the available data is still limited, and more data would be required for drawing any conclusive statements.Secondly, if sufficient participants were recruited and if the baseline of measurements of both garlic-treatment groups and control groups were comparable turned to be important factors as indications of the study quality and indeed, they had impacts on the measurement terms, such as TG and HDL, although they were not significantly changed in terms of effect sizes.
In summary, we performed an up-to-date meta-analysis to review the effects of garlic intake on CVD-associated risk factors using more detailed analyses and more systematic controls.Although a solid conclusion cannot be reached due to the limited high-quality clinical studies, our study provided valuable and reliable information for the implication of garlic in the management of risk factors of CVDs.Of interest, the effects of garlic intake on TC and LDL were more noticeable with a lower dose, and longer duration in patients with CVDs.In addition, subgroup analysis indicated that applicable diet intervention could be an important control factor.This would be an important factor to consider for any future RCTs designed to investigate the most effective dose and approach for the use of garlic in the prevention and management of CVDs.

Figure 1 .
Figure 1. a schematic flow diagram that indicates how the studies were selected.22 rCts were included in the final meta-analysis after careful filtering and selection.

Figure 2 .
Figure 2. an overview of the quality of the selected studies.

Figure 3 .
Figure 3. a detailed quality assessment of each individual study."+": low risk of bias; "−": high risk of bias and "?": unclear risk of bias.

Figure 4 .
Figure 4. Forest plots of showing the standard mean difference between the garlic treatment group and the placebo control group with the indication of 95% confidence interval for the key terms reflecting the blood lipid levels.the fixed-effect models were used for the overall effect.(a) tC: total Cholesterol; (B) ldl: low-density lipoprotein.

Figure 5 .
Figure 5. Forest plots of showing the standard mean difference between the garlic treatment group and the placebo control group with the indication of 95% confidence interval for the key terms reflecting the blood lipid levels.the fixed-effect models were used for the overall effect.(a) Hdl: High-density lipoprotein; (B) tG: triglyceride.

Figure 6 .
Figure 6.Forest plots of showing the standard mean difference between the garlic treatment group and the placebo control group with the indication of 95% confidence interval for the key terms reflecting the blood lipid levels.the fixed-effect models were used for the overall effect.(a) Homocysteine; (B) lp(a): lipoprotein (a); (C) apoa1; (d) apoB.

Figure 7 .
Figure 7. Forest plots of showing the standard mean difference between the garlic treatment group and the placebo control group with the indication of 95% confidence interval for the key terms reflecting the blood pressure.the fixed-effect models were used for the overall effect.(a) sBP: systolic Blood Pressure; (B) dBP: diastolic Blood Pressure.

Figure 8 .
Figure 8. Forest plots of showing the standard mean difference between the garlic treatment group and the placebo control group with the indication of 95% confidence interval for the key terms reflecting the liver function.the fixed-effect models were used for the overall effect.(a) alt: alanine aminotransferase; (B) ast: aspartate aminotransferase; (C) alP: alkaline Phosphatase; (d) Blood sugar.

Figure 9 .
Figure 9. Forest plots of showing the standard mean difference between the garlic treatment group and the placebo control group with the indication of 95% confidence interval for the key terms reflecting the inflammation status. the fixed-effect models were used for the overall effect.(a) tnF-a: tumor necrosis Factor alpha; (B) il-6: interleukin 6; (C) Fibrinogen; (d) CrP: C-reaction protein.

Figure 10 .
Figure 10.Meta-regression analysis between the between-group standardized mean difference (sMdbetween) and (a) study duration; (B) garlic doses given (aGe equivalent); (C) baseline values of aGe for six terms, including tC, tG, Hdl, ldl, sBP and dBP.

Figure 11 .
Figure 11.Meta-regression analysis between the between-group standardized mean difference (sMdbetween) and (d) total number of participants; (e) difference in baseline values between garlic treatment groups and placebo control groups for six terms, including tC, tG, Hdl, ldl, sBP and dBP.

Figure 13 .
Figure 13.Funnel plots showing the potential publication bias and sample size effects of the garlic intervention on changes in the 18 terms that have been assessed.the funnel plot asymmetry was tested by three approaches, including Begg, egger and Macaskill.

Table 1 .
a table that summarizes the 22 rCts that were included in the meta-analysis.

Table 2 .
Key results of the sub-group meta-analysis.

Table 3 .
Key results of the restriction and sensitivity analysis.