Background

Cardiovascular diseases (CVD), comprising ischemic heart disease, stroke, heart failure, peripheral arterial disease, and a number of other cardiac and vascular conditions are still the leading contributors to disease burden worldwide. According to the World Health Organization (WHO) by 2025, over 28 million individuals will die from CVD annually [1, 2]. There is, therefore, an urgent need to develop and implement safe and cost-effective interventions and policies [3].

Over the past decade, there has been an increasing interest in lifestyle and dietary interventions to lessen chronic disease risks. A huge body of evidence shows that fruits and vegetables intake is inversely associated with the risk of the most common chronic diseases including CVD [4]. These effects are attributed to the abundance of bioactive compounds in fruits and vegetables [5].

Anthocyanins (ACN) seem to be the major subgroup of flavonoids (pigments occurring in fruits and vegetables) having exclusive metabolic and biological behaviors [6, 7]. There are more than 700 structurally distinct ACN derivatives that naturally occur in plants as anthocyanidin (aglycone) attached to sugar moieties [8]. Important sources of ACN in a typical American diet include berries, currants, grapes, wine and vegetable sources include (red) cabbages, radishes, onions, grains, roots and tubers [9].

A growing body of evidence indicates that consumption of ACN-rich products can reduce the risk of CVD, mainly in the subjects at risk [10, 11]. For example, a clinical research demonstrated that consumption of pomegranate juice, which contains ACN and tannins, lowers blood pressure and carotid intima media thickness [12]. Also, meta-analyses of prospective cohort studies suggest that high intakes of ACN may be associated with a decreased risk of CVD [13, 14] and type 2 diabetes (T2D) [15]. Though a meta-analytical study reported the beneficial impacts of ACN on cardiometabolic health some five years ago [16], current data are still conflicting [17, 18]. Furthermore, the possible effects of ACN on some important variables including waist circumference and body fat mass have not already been summarized and analyzed. In this meta-analysis, findings from recent studies on the preventive properties of ACN against CVD are summarized and current knowledge of the field is updated. To do this, we included 41 clinical trial studies and performed subgroup analysis and meta-regression for the cardiometabolic risk factors including body mass index, waist circumference and body fat mass, that have not been examined in the previous meta-analysis.

Methods

This systematic review and meta-analysis was planned, conducted, and reported according to the widely accepted quality standards [19] and was registered at http://www.crd.york.ac.uk/Prospero (Registration no. CRD42021286466).

Types of interventions and outcome measures

The interventions investigated in this review were giving ACN either as supplement or food in various doses and frequencies as compared with a control group. No limit was placed on the dose, type of vehicle or frequency at which supplements and/or fortified foods were taken.

Outcome measures were cardiometabolic risk factors including:

Anthropometric measures: weight (kg), body mass index (BMI) (kg/m2), waist circumference (WC) (cm), body fat mass (FM)(%).

Glycemic indicators: fasting blood glucose (FBG) (mg/dL), fasting serum insulin (FSI) (µIU/mL), two-hour glucose tolerance test (mg/dL), homeostasis model assessment-estimated insulin resistance (HOMA-IR), glycated hemoglobin (HbA1c) (%).

Lipid profile: serum triglyceride (TG) (mg/dL), total cholesterol (TC) (mg/dL), low density lipoprotein-cholesterol (LDL-C) (mg/dL), high density lipoprotein-cholesterol (HDL-C) (mg/dL), apolipoprotein A (Apo A) (mg/dL), apolipoprotein B (Apo B) (mg/dL).

Search strategy

A comprehensive search of electronic databases was conducted for eligible trials up to September 2021. The following databases were covered: National Library of Medicine (PubMed), Scopus, Web of Science (WoS), and Cochrane Database of Systematic Reviews (Cochrane Library, CDSR) using the following search terms in titles and abstracts: 1- (anthocyanin) AND (Anthropometric measures: weight, body mass index (BMI), waist circumference (WC), fat mass (FM)) 2- (anthocyanin) AND (Glycemic indicators: fasting blood glucose (FBG), fasting serum insulin (FSI), two-hour glucose tolerance test, homeostasis model assessment-estimated insulin resistance (HOMA-IR), HbA1c), 3- (anthocyanin) AND (lipid profile: serum triglyceride (TG), total cholesterol (TC), low density lipoprotein-cholesterol (LDL-C), high density lipoprotein-cholesterol (HDL-C), apolipoprotein A (apo A), apolipoprotein B (apo B)).

All of the studies were limited to English language and those in humans. All the references in the papers of the included studies were scanned for further relevant reports.

Types of studies: inclusion and exclusion criteria

Two investigators independently searched and assessed articles for eligibility according to the following inclusion criteria:

  1. (1)

    all studies with a randomized clinical trial design;

  2. (2)

    data description as mean ± SD and sufficient data of outcomes at both baseline and the end of intervention in each group or the net change values provided

Studies were excluded based on the following criteria:

  1. (1)

    incomplete data;

  2. (2)

    duplicate publication of articles;

  3. (3)

    obscurely reported outcomes, or lack of control groups; and

  4. (4)

    animal as well as non-interventional studies

Data extraction and management

Two reviewing authors independently screened titles or abstracts to exclude studies that failed to meet the inclusion criteria and then obtained the full-text reports for further evaluations. Discrepancies were resolved through consensus. Detailed data of study design, context, participants’ information, interventions and outcomes were extracted. For multi-armed studies, pairs of arms relevant to the review were compared. Data of the control group were used for each intervention group comparison. The weight assigned to the control group was reduced by dividing the control group number (N) by the number of intervention groups. The selection process in sufficient detail to complete a PRISMA flow diagram was recorded in the supplementary data.

Assessment of risk of bias in the included studies

In order to assess the risk of bias for each of the included studies, the two reviewing authors independently assessed the quality of the included studies according to the Cochrane risk of bias tool [20] and following criteria: (i) random sequence generation, (ii) allocation concealment, (iii) blinding of participants, (iv) blinding of personnel and outcome assessment, (v) incomplete outcome data, and (iv) selective outcome reporting and other bias (bias due to problems not covered elsewhere, e.g. industry funding).

Assessment of reporting biases

Funnel plots were used to assess the potential existence of bias (were not shown). We performed a regression asymmetry test for detection of bias [21]. In this study, p < 0.05 was considered significant.

Subgroup analysis and investigation of heterogeneity

If we identified substantial heterogeneity and there were sufficient data, we would investigate it using subgroup as well as sensitivity analyses. The following subgroup analyses were carried out:

  • participants’ health status ( having diabetes, dyslipidemia, …)

  • type of intervention (supplements/extract vs. food)

Sensitivity analysis

The aim of the sensitivity analysis was to examine the effect size when some studies had less stringent inclusion criteria. If there were sufficient data to allow for sensitivity analysis, this analysis would be performed for examining the effects on results by excluding:

  • (i) Trials at high risk of bias, as specified in the Assessment of risk of bias in included studies section.

  • (ii) Trials with small sample sizes (less than 15 participants in each group);

Moreover, the leave-one-out analysis was performed to assess the key studies with substantial impact on between-study heterogeneity.

Meta-regression

Random‐effects model meta‐regression analyses were performed to assess sources of inter-trial heterogeneity. The follow-up time, ACN dose and variables at baseline (according to model and dependent variable) were considered as covariates.

Statistical analyses

Meta-analysis was performed using STATA version 17.0 (StataCorp, College Station, TX). For continuous outcomes, a mean difference (MD) and 95% confidence interval (CI) calculated for each study (i.e. intervention group minus control group differences).

The data reported as median were converted to mean and SD using the formula recommended by the Cochrane Collaboration Handbook [20].

In addition, heterogeneity was assessed using Q and I2 tests. The fixed effect model was used as pooling method when there was no statistically significant heterogeneity (I2 < 50%), whereas the random-effects model was employed on the contrary (I2 > 50%).

The Egger’s test was used to assess risk of publication bias.

Results

Initially, 2,512 studies were found in a preliminary search from which 2,444 studies were excluded after removing duplicates and primary screening of titles and abstracts, another 21 studies were excluded due to following reasons:

  • 1) There was no control group;

  • 2) The studies provided no available data.

Finally, a total of 47 studies were selected for the analysis. A flow diagram showing the selection process is presented in Fig. 1.

Fig. 1
figure 1

The flowchart of the study protocol on the effects of anthocyanins intake on cardiometabolic risk factors

Full size image

Characteristics and bias risk assessment of included studies

The basic characteristics of the 47 randomized clinical trials (RCTs) are listed in Table 1. There were eight studies in healthy participants [22,23,24,25,26,27,28,29], four in prediabetic adults [30,31,32,33], eight in subjects with metabolic syndrome [34,35,36,37,38,39,40,41], one in elderly with mild to moderate dementia Alzheimer [42], four in overweight/obese subjects [43,44,45,46], one in adults with cardiovascular risk factors [47], seven in patients with T2D [48,49,50,51,52,53,54], eight in dyslipidemia [55,56,57,58,59,60,61,62], one in subjects with non-alcoholic fatty liver disease [63] and three in patients with ischemic coronary heart disease [64,65,66]. Trial intervention duration ranged from 3 weeks [46] to 24 months [50]. The sample size from all the trials ranged between 20 [29, 54, 67] and 160 participants [32, 33].

Table 1 Characteristics of the studies selected for meta-analysis of the effects of anthocyanins on cardiometabolic risk factors
Full size table

A total of 23 trials compared provision of ACN supplement/extract in comparison with placebo or no intervention [23, 25, 28, 31,32,33, 43, 44, 46, 48,49,50,51,52, 54,55,56,57,58,59,60,61,62,63, 65].

Risk of bias in included studies

All trials provided data on losses to follow-up; eleven studies reported losses of > 10% [28, 30, 31, 34, 36, 40, 42, 44, 48, 55, 65] (Additional file 1: Figures S1 and S2). In thirteen trials, the information on the methods used for allocation concealment was unclear [23, 24, 30, 34,35,36,37,38, 42, 43, 47, 67, 68]. Three trials reported small sample size (< 15 subjects in each arm) [27, 67, 68].

Effect of anthocyanin on body mass index

Changes of BMI (kg/m2) after intervention with ACN were evaluated in twenty studies. From two studies that had multi-arm intervention, two pairs of arms were included in the analysis [30, 37]. Four studies included participants who were overweight or obese at baseline [27, 43,44,45].

The results of twenty trials showed that intervention with ACN resulted in a significant decrease in BMI (1,280 participants, mean difference (MD),− 0.21; 95% CI, − 0.38, − 0.04; P < 0.001) (Additional file 1: Figure S3).

The meta-regression analysis (random effects and maximum likelihood) with the difference of means as the dependent variable and the follow-up time (β = 0.01; 95% CI − 0.02, 0.042; p = 0.484), ACN dose (β = 0.00; 95% CI − 0.001, 0.00; p = 0.466) and BMI at baseline (β = 0.23; 95% CI − 0.04, 0.09; p = 0.490) as covariates showed no significant effects.

After conducting stratified analyses according to the type of ACN (supplement/extract vs. food), we found a statistically significant effect in both subgroups (Fig. 2). However, unlike supplement (MD: − 0.35; 95% CI − 0.47, − 0.23; P < 0.001), using ACN as powdered or food was accompanied by an increase in BMI in the intervention group as compared with control (MD: 0.1, 95% CI 0.04, 0.16; P = 0.001).

Fig. 2
figure 2

Forest plot of subgroup analysis of the effect of anthocyanins compared with control on body mass index (kg/m2)

Full size image

Sensitivity analyses were performed and made no remarkable change in overall results (MD: − 0.20, 95% CI − 0.39, − 0.04, P = 0.01). The results of leave-one-out analysis were shown in the (Additional file 1: Figure S4).

The Egger’s tests indicated no significant publication bias (P = 0.605).

Effect of anthocyanin on waist circumference

Eight trials compared the effect of ACN intake, as supplement/foods, versus control on waist circumference (WC). From one trial, two paired-arms were included in the analysis. Pooled data showed no statistically significant effect on WC (614 participants, MD: 0.78 cm, 95% CI − 0.43, 1.98 cm; P = 0.21) (Additional file 1: Figure S5).

Sensitivity analysis was performed for examining the effects of excluding trials with small sample sizes. The results were shown no significant effect (MD: − 0.09 cm, 95% CI − 1.25, 1.07; P = 0.88).

The results of leave-one-out analysis were shown in the (Additional file 1: Figure S6).

Effect of anthocyanin on body fat mass

Four trials compared the effect of ACN intervention with control on body fat mass (FM). Pooled data showed statistically significant effect on FM changes (150 participants, MD: − 0.3%, 95% CI − 0.42 to − 0.18%, p < 0.001) (Additional file 1: Figure S7).

The meta-regression analysis (random effects and maximum likelihood) with the difference of means as the dependent variable and the follow-up time (β = 0.01; 95% CI − 0.24, 0.27, p = 0.916) and used dose of ACN (β = 0.001; 95% CI − 0.01, 0.014, p = 0.853) as covariates showed the insignificant effects.

There were no data to allow for sensitivity analysis. The Egger’s tests indicated no significant publication bias (P = 0.881).

The results of leave-one-out analysis were shown in the (Additional file 1: Figure S8).

Effect of anthocyanin on fasting blood glucose

We pooled and analyzed data from 28 RCTs evaluating the effects of ACN intake on FBG. From three studies that had multi-arm intervention, two pairs of arms were included in the analysis [30, 37]. The forest plots for the effects of ACN intake on FBG are shown in the Additional file 1: Figure S9. The pooled MD was − 2.55 mg/dL (95%CI − 4.23, − 0.87), indicating a significantly greater reduction in FBG concentrations in the ACN intervention group than in the control (placebo) group.

The subgroup analyses were carried out to explore the impact of type of intervention (supplement/extract vs. food) and health status of participants (with diabetes vs. without diabetes). Significant reductions were observed in both subgroups of T2D patients (434 participants, MD: − 15.33 mg/dL; 95% CI − 25.8, − 7.86 mg/dL; P < 0.001) and participants without diabetes (1124 participants, MD: − 1.55 mg/dL; 95% CI − 2.38, − 0.72 mg/dL, P = 0.001). However, reduction of serum glucose concentrations was significantly more in the subjects with T2D (p < 0.001).

After a subgroup analysis, we found that FBG decreased significantly only in those who used ACN as supplement/extract (MD: − 3.13 mg/dL, 95% CI − 4.9, − 1.35 mg/dL; P = 0.004) (Fig. 3).

Fig. 3
figure 3

Forest plot of subgroup analysis of the effect of anthocyanins compared with control on fasting blood glucose (mg/dL)

Full size image

The univariate meta-regression suggested that a higher baseline FBG was associated with a better effect of ACN intake on FBG changes (β = − 0.25; 95% CI − 0.34, − 0.15; P < 0.001) and this association remained significant after adjusting for duration of intervention, age of participants, dose of ACN and baseline BMI (β = − 0.25; 95% CI − 0.37, − 0.12; P < 0.001).

Sensitivity analyses were performed to test the robustness of the results and made no remarkable change in overall results (MD: − 3.0 mg/dL, 95%CI − 5.6, − 0.4 mg/dL, P = 0.02).

The results of leave-one-out analysis were shown in the (Additional file 1: Figure S10).

The Egger’s tests suggested that no significant publication bias (P = 0.089).

Effect of anthocyanin on fasting serum insulin

Eleven randomized controlled trials (two paired-arms from two studies) (n = 653) had measured fasting insulin. Results showed no significant effect of ACN intake on fasting insulin (MD: 0.64 µIU/mL; 95% CI − 0.33, 1.61 µIU/mL; P = 0.19) (Additional file 1: Figure S11).

A subgroup analysis was conducted to investigate heterogeneity using known factors that might influence circulating insulin, including participants (with vs. without diabetes) and intervention type (supplement/extract vs. food). Analyses did not show any significant influence (Additional file 1: Figure S12).

Results of sensitivity analysis showed no significant effect of ACN intake on fasting insulin (MD: 0.064 µIU/mL; 95% CI − 92, 1.09 µIU/mL; P = 0.87). The regression asymmetry test was not significant (p = 0.413).

The results of leave-one-out analysis were shown in the supplementary file (Additional file 1: Figure S13).

Effect of anthocyanin on HbA1c

There were ten studies (three paired-arms from two studies) that had examined the effects of ACN intake on HbA1c. Meta-analysis results indicated that ACN consumption decreased HbA1c (MD: − 0.14%; 95% CI − 0.26, − 0.02%; P = 0.03). However, subgroup analysis showed this effect only in the participants with T2D (MD: − 0.29%; 95% CI − 0.47, − 0.10%; P = 0.003) and in those trials that used ACN as supplement/extract (MD: − 0.17%; 95% CI − 0.27, − 0.07; P = 0.001) (Fig. 4, and Additional file 1: Figure S14).

Fig. 4
figure 4

Forest plot of subgroup analysis of the effect of anthocyanins compared with control on HbA1c (%)

Full size image

The univariate meta-regression suggested that a higher baseline FBG was associated with a better effect of ACN intake on HbA1c changes (β = − 0.13; 95% CI − 0.24, − 0.03; p = 0.011) and this association remained significant after adjusting for ACN dose and baseline BMI (β = − 0.16; 95% CI − 0.31, − 0.019; p = 0.026).

Sensitivity analyses were performed to test the robustness of the results and revealed no remarkable change in overall results (MD: − 0.19%, 95%CI − 0.4, − 0.02%, P = 0.08). The Egger’s tests indicated no significant publication bias (P = 0.532).

The results of leave-one-out analysis were shown in the (Additional file 1: Figure S15).

Effect of anthocyanin on serum triglycerides concentrations

We pooled and analyzed data from 38 RCTs evaluating the effects of ACN intake on serum TG concentration. From Four studies [30, 32, 37, 59] that had multi-arm intervention, two pairs of arms were included in the analysis (Additional file 1: Figure S16). The pooled MD was − 9.46 mg/dL (95%CI, − 14.6, − 4.31 mg/dL; P < 0.001, indicating a significantly greater reduction in TG concentrations in participants who received ACN than those used placebo.

The subgroup analysis test showed that there was a significant effect of ACN on TG concentrations in all subgroups of participants (with vs. without dyslipidemia at baseline) and intervention type (supplement/extract vs. food). Furthermore, there were no significant differences between subgroups (p = 0.27 and p = 0.13, respectively) (Additional file 1: Figure S17).

In meta-regression, changes in serum TG concentrations following ACN intervention were found to be independent of treatment duration (β = − 0.49; 95% CI − 2.17, 1.18; p = 0.566), baseline BMI (β = − 0.56; 95% CI − 2.3, 1.18; p = 528) and age of subjects (β = 0.59; 95% CI − 0.39, 1.58; p = 0.239). Nevertheless, changes in TG were significantly associated with the baseline TG concentrations (β = − 0.21; 95% CI − 0.29, − 0.12; p < 0.001).

Results of sensitivity analysis showed significant effect of ACN intake on serum TG similar to the overall effect (MD: − 11.68 mg/dL; 95% CI − 18.04, − 5.32 mg/dL; P < 0.001). The regression asymmetry test was not significant (p = 0.971).

The results of leave-one-out analysis were shown in the (Additional file 1: Figure S18).

Effect of anthocyanin on total cholesterol concentrations

Data from 40 trials (four trials with multi-arm interventions) involving 2433 subjects showed that consumption of ACN, as compared with placebo, reduces serum TC concentration (MD: − 6.98%; 95% CI − 11.97, − 2.0 mg/dL; P = 0.01) (Additional file 1: Figure S19).

Subgroup analyses were carried out for participants (with vs. without dyslipidemia at baseline) and intervention (supplement/extract vs. food). The mean change in TC concentration in trials using supplement/extract (MD: − 9.76 mg/dL, 95% CI − 17.33 to − 2.2 mg/dL; P = 0.011) was greater than from daily foods (MD: − 3.10 mg/dL; 95% CI − 8.24, 2.05 mg/dL; P = 0.238). Also, studies with baseline dyslipidemia showed greater decreases in TC concentrations (MD: − 13.41 mg/dL; 95% CI − 22.95, − 3.88 mg/dL; P = 0.006) than those without baseline dyslipidemia (MD: − 2.30 mg/dL; 95% CI − 6.78, 2.18 mg/dL; P = 0.315) (Additional file 1: Figure S20).

The univariate meta-regression suggested that a higher baseline TC was associated with a better effect of ACN intake on TC changes (β = − 0.36; 95% CI − 0.60, − 0.15; P = 0.002). Adjustments for ACN dose (β = 0.017; 95% CI − 0.02, 0.052; P = 0.327), duration of intervention (β = 1.12; 95% CI − 0.19, 1.44; P = 0.853) and baseline BMI (β = − 0.36; 95% CI − 1.71, 0.98; P = 0.594) revealed no significant effect of these variables.

Sensitivity analyses were performed to test the robustness of the results and showed effects similar to the overall effects (MD: − 8.63 mg/dL, 95%CI − 14.6, − 2.61, P < 0.001). The regression asymmetry tests suggested that no significant publication bias (P = 0.200).

The results of leave-one-out analysis were shown in supplementary file (Additional file 1: Figure S21).

Effect of anthocyanin on LDL-C concentrations

Thirty nine trials compared the effect of ACN intervention on serum LDL-C concentrations with control. Pooled data showed statistically significant effect on LDL-C changes (MD: − 6.91 mg/dL; 95% CI − 9.99, − 3.83 mg/dL; P < 0.001) (Additional file 1: Figure S22).

The effect of ACN was similar in all types of intervention (supplement/extract vs. food) (Additional file 1: Figure S23). Nevertheless, ACN intake could decrease serum LDL-C concentrations only in dyslipidemic subjects (MD: − 10.68 mg/dL; 95% CI − 14.85, − 6.52 mg/dL; P < 0.001) but not in those with normol blood lipids (MD: − 3.77 mg/dL; 95% CI − 7.6, 0.06: P = 0.054).

Our meta-regression suggested the association between ACN intake and LDL-C was not affected by duration of intrvention (β = − 0.25; 95% CI − 1.03, 0.53; P = 0.531), dose of intervention (β = − 0.003; 95% CI − 0.028, 0.021; P = 0.786), BMI at baseline (β = − 0.56; 95% CI − 1.54, 0.41; P = 0.257) and LDL-C concentrations at baseline (β = − 0.12; 95% CI − 0.32, 0.07; P = 0.216).

Results of sensitivity analysis showed significant effect of ACN intake on serum LDL-C MD:—7.72 mg/dL; 95% CI − 11.23, − 4.21 mg/dL; P < 0.001). The regression asymmetry test was not significant (p = 0.365).

The results of leave-one-out analysis were shown in (Additional file 1: Figure S24).

Effect of anthocyanin on serum HDL-C concentrations

Serum HDL-C concentration was evaluated in 39 studies (four studies with multiple-arms). Meta-analysis revealed positive effect of ACN intake on HDL-C concentrations (MD: 1.71 mg/dL; 95% CI 0.43, 2.08 mg/dL; p = 0.01) (Additional file 1: Figure S25).

Subgroup analyses were carried out for participants (with vs. without dyslipidemia at baseline) and intervention (supplement/extract vs. food). The mean change in serum HDL-C concentration in those trials on subjects with dyslipidemia (MD: 2.35 mg/dL; 95% CI 0.26, 4.45 mg/dL; P = 0.028) was greater than the trials conducted on participants normal blood lipids (MD: 0.73 mg/dL; 95% CI 0.03, 1.43 mg/dL; P = 0.041) (Additional file 1: Figure S26).

Findings from meta-regression suggested the association between ACN intake and serum HDL-C was not affected by duration of intervention (β = 0.26; 95% CI − 0.06, 0.58; P = 0.114), dose of intervention (β = − 0.002; 95% CI − 0.01, 0.005; P = 0.494), BMI at baseline (β = − 0.0008; 95% CI − 0.34, 0.34; P = 0.996) and serum HDL-C concentrations at baseline (β = − 0.04; 95% CI − 0.09, 0.012; P = 0.130).

Sensitivity analyses were performed to test the robustness of the results and showed positive effect of ACN intake on HDL-C concentrations similar to the overall effects (MD: − 2.16 mg/dL, 95% CI 0.71, 3.6 mg/dL, P < 0.001). The Egger’s tests indicated no significant publication bias (P = 0.182).

The results of leave-one-out analysis were shown in (Additional file 1: Figure S27).

Effect of anthocyanin on serum apolipoprotein A (Apo A) concentrations

Meta-analysis of RCTs yielded a weighted estimate for the overall effect of ACN intake of MD: 3.94 mg/dL (95% CI 0.77 to 7.11) on Apo A concentrations (Additional file 1: Fig. S28). Apo A estimates were not significantly different among subgroups (without dyslipidemia vs. with dyslipidemia and supplement/extract vs. food) (Additional file 1: Figure S29).

The meta-regression analysis suggested that effect of the ACN intake on Apo A concentrations was not affected by duration of intrvention (β = − 0.49; 95% CI − 2.5, 1.51; P = 0.631), dose of intervention (β = 0.0005; 95% CI − 0.02, 0.021; P = 0.962), BMI at baseline (β = − 0.71; 95% CI − 2.76, 1.32; P = 0.490) and Apo A concentrations at baseline (β = 0.18; 95% CI − 0.06, 0.42; P = 0.155).

Results of sensitivity analysis showed similar significant effect of ACN intake on serum apo A, (MD: 3.91 mg/dL; 95% CI 0.16, 7.65 mg/dL; P = 0.04). The regression asymmetry test was not significant (p = 0.301).

The results of leave-one-out analysis were shown in (Additional file 1: Figure S30).

Effect of anthocyanin on serum apolipoprotein B (Apo B) concentrations

The meta-analysis showed that ACN intake was not associated with significant changes in serum apo B concentrations (MD: − 2.36 mg/dL; 95% CI − 5.55, 0.83 mg/dL; P = 0.15). There was no statistically significant changes in circulating apo B in subgroup analysis, either (Additional file 1: Figures S31 and S32).

The meta-regression analysis suggested that effect of ACN intake on serum apo B concentrations was not affected by duration of intervention (β = − 1.13; 95% CI − 3.64, 1.38; P = 0.378), dose of intervention (β = − 0.015; 95% CI − 0.04, 0.009; P = 0.223), BMI at baseline (β = 3.6; 95% CI − 3.02, 10.22; P = 0.287) and apo B concentrations at baseline (β = 0.39; 95% CI − 0.26, 1.05; P = 0.241).

Results of sensitivity analysis showed no significant effect of ACN intake on serum apo B, (MD: − 1.9 mg/dL; 95% CI − 5.27, 1.47 mg/dL; P = 0.27). The regression asymmetry test was not significant, either (p = 0.128).

The results of leave-one-out analysis were shown in (Additional file 1: Figure S33).

Discussion

The findings of this meta-analysis confirmed the positive effects of ACN intake on certain cardiometabolic risk factors including BMI and more importantly FM, glycemic and lipidemic profiles. A combination of these factors occur in metabolic syndrome (MetS) that is characterized by central obesity, dysglycemia, above normal blood pressure, low serum HDL-C and high TG concentrations [69]. There is a remarkable higher risk of developing various diseases including CVD and diabetes in the subjects with MetS and that is why prevention and treatment of MetS can actually reduce CVD risk [70].

Anthocyanins and body mass index

We found the positive effect of ACN consumption on reduction of BMI which further confirms previous reports [71, 72]. Nevertheless, our finding on the supplement form-dependent effect of ACN consumption on BMI is quite noticeable, i.e. ACN supplement (like tablet) causes decrease whereas powdered ACN results in increase in BMI. On re-analysis of the data, however, we found out that this controversial finding was mostly due to one study [41]. Notwithstanding, in that study the increase in body weight and hence BMI was mostly due to lean body mass and not fat mass [41]. When we excluded this study from analysis, this effect was removed (MD − 0.14 (− 0.66, 0.9), P = 0.610).

Sensitivity analysis using the leave-one-out approach indicated that the direction of the combined estimates did not vary markedly with the removal of each study in turn. However, removal of Stull et al. [41] that used yogurt and skim milk-based smoothie as intervention, changed the effect size from − 0.21 to − 0.35.

Anthocyanins and body fat mass

Some evidence indicates that truncal adiposity is the core constituent of the metabolic derrangements observed in MetS [73, 74]. Our data showed the positive effect of ACN intake on total FM and not on central obesity, as judged by WC. However, we were unable to perform subgroub analysis due to the limited eligible RCTs.

The association between body FM variability and CVD risk has already been reported [75]. Nonetheless, strong evidence supports the contribution of truncal fat distribution, even in the context of normal BMI, to the increased risk of CVD [76, 77]. One possible explanation for this dicrepancy in results could be the difference in precision of the methods of two assessements as FM is assessed using an instrument like dual-energy X-ray absorptiometry (DEXA) whereas WC is measured using a measuring tape that is less sensitive to detect changes. This is especially the case when changes occur in visceral fat [78] that is more related to CVD risk [79].

The exact mechanism of anti-adipogenesis function of ACN has not been fully elucidated yet. Some evidence suggests an inhibitory effect of ACN against adipocyte differentiation that might be exerted through 5' adenosine monophosphate-activated protein kinase (AMPK) signaling pathway activation [80, 81].

Anthocyanins and glycemic status

The findings of this meta-analysis confirmed the alleviative effect of ACN on glycemic status including FBS and HbA1c. Notwithstanding, the effect on both variables was more prominent in those subjects with above-normal values, notably in the subjects with T2D. It has been documented that ACN has the potential to upregulate and also downregulate certain glucose metabolic pathways including AMPK whereby exerts its modulating effect on glucose homeostasis [82]. Other proposed mechanisms for anti-diabetic actions of ACN include inhibition of α-amylase, α-glucosidase and modulating effect on glucose transporters, inhibition of glycogenolysis and modulation of gut microbiota [83].

We found better effect on HbA1c of ACN when consumed as a supplement. It has been shown that ACN in foods may interact with other biomolecules including proteins and polysaccharides that may affect ACN activity [83]. Evaluation of the efficacy of ACN in natural foods versus supplements need further investigations.

Anthocyanins and lipidemic status

Our findings documented the effects of ACN on healthy changes of blood lipids including a decrement of TG, TC and LDL-C and an increment of HDL-C. Again here, better effects were achieved in the subjects with deranged blood lipid profile. Nevertheless, data from this meta-analysis failed to show any effect of ACN on serum concentrations of apo A and apo B. The mechanism of action of ACN on healthy modification of blood lipids is under investigation. Some studies indicate that ACN can reduce TC possibly via increased fecal excretion of neutral and acidic sterols together with down-regulation of hepatic β-hydroxy β-methyl glutaryl coenzyme A (HMGCoA) reductase, the rate-limiting enzyme of cholesterol biosynthesis [84]. The mechanism of TG-lowering effect of ACN remains to be elucidated. This effect has been attributed to the reduction of apo B and CIII [85]. Notwithstanding, our meta-analysis failed to show any significant effect of ACN on serum apo B concentrations. ACN-induced decrement of circulating LDL-C may be partly due to downregulation of cholesteryl ester transfer protein [86] and upregulation of LDL-receptor [87]. It has also been proposed that ACN can affect both liver and blood lipids contents via activation of AMPK. Activated AMPK in turn induces acetyl coenzyme A carboxylase phosphorylation and upregulates hepatic peroxisome proliferator-activated receptor-alpha (PPAR-α), acylCoA oxidase and palmitoyltrasferase-1 A [88].

Limitations

This study has some limitations. We did not evaluate the antioxidant effect of ACN that can have a major contribution to prevention of CVD [89, 90] and other diseases [91, 92]. Nonetheless, the efficacy of ACN as an antioxidant has been recently confirmed in a meta-analysis [93]. Limited studies on effects of ACN intake on waist circumference, fat mass, apo A and apo B concentrations (n < 10) could effect on generalizability of results of meta-regression tests. Furthermore, we did not evaluate the effect of ACN on biomarkers of systemic and vascular inflammation, which has an axial role in both development and progression of CVD [94, 95].

The other limitation of our study is scanty available RCTs that evaluated the long-term effects of ACN intake. Though some RCTs had treatment durations of up to two years, the question of the effectiveness of long-term ACN intake remains unanswered due to the lack of experimental evidence.

In some RCTs, especially those that used ACN-containing foods, the exact amount of ACN consumption was not reported. Tt was not, therefore, possible to perform dose–response analysis. Besides, in most RCTs, ACN intake from dietary sources was not adjusted and this could affect the results.

The intervention strategy across the RCTs was diverse with respect to the dosage of ACN, the frequency of intake and type of intervention (food/supplement), which inevitably contributed to the high levels of heterogeneity. We attempted to reduce the heterogeneity by choosing appropriate statistical models and using sensitivity analyses. Notwithstanding, this point should be considered when interpreting and generalizing the results.

Finally, the present meta-analysis employed just the papers published in English. Thus, the papers in other languages or unpublished data might potentially increase the heterogeneity in the present results.

Conclusion

The findings of this meta-analysis indicate that ACN intake in the forms of natural foods and supplements can induce healthy changes in body FM, glycemic and lipidemic status and these effects are more prominent in the subjects with above-normal values. Further well-designed studies are needed to evaluate and compare the efficacy of ACN consumption as natural food sources versus supplements. Meanwhile, fortification of foods and drinks with ACN can also be considered to increase ACN intake in the general population as a preventive strategy against many diseases notably CVD.