The effect of anthocyanins supplementation on liver enzymes: A systematic review and meta‐analysis of randomized clinical trials

Abstract This systematic review and meta‐analysis aimed to assess effect of consuming anthocyanins (ACNs; pure ACNs or products containing ACNs) on liver enzymes levels including alanine aminotransferase (ALT), aspartate aminotransferase (AST), and gamma‐glutamyl transferase (GGT). Although no significant impact was detected on the liver enzymes, a significant reduction was observed on ALT (WMD = −4.932 U/L, 95% CI = −9.848 to −0.015, p = .049) and AST (WMD = −3.464 U/L, 95% CI = −6.034 to −0.894, p = .008) in the studies that examined them as primary outcomes. A significant decrease was found on AST among the healthy subjects (WMD = −4.325 U/L, 95% CI = −8.516 to −0.134, p = .043) and in the studies that used products containing ACNs as intervention (WMD = −2.201 U/L, 95% CI = −4.275 to −0.127, p = .037). Although no significant relation was detected between ACNs dosage and the liver enzymes, significant associations were found between the duration of trial with ALT (ALT: slope: 0.09, 95% CI = 0.040 to 0.139, p = .0003) and AST (slope: 0.076, 95% CI = 0.037 to 0.115, p = .0001). In conclusion, although ACNs had no significant effect on the liver enzymes, a significant decrease was discovered on ALT and AST in the studies that evaluated them as primary outcomes. A significant reduction was observed in AST in the healthy individuals and in the studies used products containing ACNs as intervention. Significant relations were also found between the duration of trial with ALT and AST. Further studies are required to confirm these results.


| INTRODUC TI ON
Anthocyanins (ACNs) are a group of water-soluble natural pigments.
They belong to flavonoids and are found in different plant sources such as fruits, vegetables, grains, and cereals in red, purple, and blue (Bueno et al., 2012;Sangsefidi et al., 2018). Evidences indicated beneficial impacts of ACNs on disorders involving inflammatory and oxidative processes such as cardiovascular diseases (Wallace et al., 2016), type 2 diabetes (Sancho & Pastore, 2012), metabolic syndrome (Tsuda, 2008), and dyslipidemia (Liu et al., 2016). In addition, useful effects of these components were reported on glycemic control (Daneshzad et al., 2019; and lipid profile (Daneshzad et al., 2019;Shah & Shah, 2018;. Based on the literature, increase of the serum liver enzymes can be related to the injury of liver cells due to some factors including oxidative stress, and inflammation or increased fat storage in the liver following insulin resistance (Ahn et al., 2014;Bonnet et al., 2011;Suda et al., 2008;Zhang et al., 2010). The results of some studies indicated that ACNs had beneficial impacts on liver disorders such as NAFLD, and borderline hepatitis, as well as liver enzymes (Chang et al., 2014;Oki et al., 2016;Suda et al., 2008;Zhang et al., 2015).
The protective effects of ACNs on liver were attributed to antiinflammatory and antioxidant roles of these compounds, improvement of insulin resistance, lipid profile, and glycemic control (Chang et al., 2014;Guo et al., 2014;Oki et al., 2016;Suda et al., 2008;Zhang et al., 2015). However, findings of surveys over the effect of ACNs on liver enzymes levels are inconsistent and controversial. For example, in a research on NAFLD patients, supplementation with purified ACNs for 12 weeks decreased the alanine aminotransferase (ALT) levels (Zhang et al., 2015). In addition, consuming the purple sweet potato extract beverage for 8 weeks was related to decreased liver enzymes levels among healthy Caucasians with borderline hepatitis (Oki et al., 2016). Similarly, a significant reduction was found in concentration of liver enzymes following the consumption of a sweet potato extract beverage for 12 weeks in healthy men with borderline hepatitis (Suda et al., 2008). Consumption of freeze-dried blueberries for 8 weeks was associated with reduced concentrations of ALT and aspartate aminotransferase (AST) in men with type 2 diabetes (Stote et al., 2020). However, some other studies showed that ACNs had no significant effect on liver enzyme; for instance, intake of an elderberry extract in postmenopausal healthy women (Curtis et al., 2009) and pure ACNs in patients with pre-diabetes  did not have any significant effect on liver enzymes. Moreover, ingestion of Hibiscus sabdariffa extract did not affect liver enzymes among adults with NAFLD significantly (Chang et al., 2014). To the best of our knowledge, no systematic review and meta-analysis have ever been carried out on this issue. Therefore, the present systematic review and meta-analysis were conducted to prepare a more accurate estimate of the overall effect of ACNs on liver enzymes. Our objective was to investigate the impact of supplementation with ACNs (pure ACNs or products rich in ACNs) on liver enzymes levels including alanine aminotransferase (ALT), aspartate aminotransferase (AST), and gamma-glutamyl transferase (GGT).

| Search strategy
The current systematic review and meta-analysis were carried out according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Guidelines (PRISMA) (Moher et al., 2015). This study was registered at crd.york.ac.uk/Prospero as CRD42020150700.
Several databases including PubMed, ISI Web of Science, Scopus, and Google Scholar were searched up to 25 Jun 2020 without any restrictions to identify eligible researches. Medical Subject Heading (MeSH) terms and non-MeSH terms were applied to evaluate the impact of ACNs on liver enzymes levels. The following keywords were used to search: ((anthocyanins OR "anthocyanin extract" OR cyanidin OR pelargonidin OR pelargonidin OR delphinidin OR peonidin OR petunidin) AND (liver OR "liver enzyme" OR *transaminase* OR *aminotransferase* OR *transpeptidase* OR "alanine transaminase OR "alanine aminotransferase" OR ALT OR SGPT OR "aspartate aminotransferases" OR transaminases OR AST OR SGOT OR "alkaline phosphatase" OR ALP OR "gamma-glutamyltransferase" OR "gamma glutamyltransferase" OR GGT OR "gammaglutamyltransferase" OR "lactate dehydrogenase" OR "L-lactate dehydrogenase" OR "dehydrogenase L-lactate" OR "dehydrogenase lactate" OR LDH OR "ASTto-ALT ratio" OR "ALT to AST ratio" OR "liver enzyme abnormality" OR "liver enzyme activity" OR "liver function tests" OR LEA OR "AST/ALT")).
Furthermore, to ensure about the comprehensiveness of searches, references of the included studies were checked for further possible sources.

| Selection criteria
The selected surveys had the following criteria: (a) had RCT design; (b) assessed the effect of pure ACNs or products rich in ACNs including extracts, beverages, powders or juices on liver enzymes levels versus placebo/control; (c) presented the administered ACNs dosage or reported a quantifiable ACNs content for products rich in ACNs; (d) included participants aged ≥18 years; and (e) reported sufficient data for liver enzymes levels. We excluded studies if they had additional intervention other than pure ACNs or products rich in ACNs such as additional supplement or additional herbal products.

| Study selection
The initial screening was conducted by two independent researchers (ZS.S and S.K-H), who studied the articles' titles and abstracts.
Then, the full texts of all related trials were evaluated via reviewers to select the articles about the effect of ACNs (pure ACNs or products rich in ACNs including extracts, beverages, powders or juices) on liver enzymes. Eventually, any possible disagreement was negotiated and solved by consultation with other researchers (M.H and H.M-K; Figure 1).

| Data extraction
Data were extracted from the selected studies by the following criteria: authors' family names; publication year; sample size; loss to follow-up; intervention type and its dosage; study duration; crossover or parallel study design; participants' gender, age, and health status; mean and standard deviation (SD) of liver enzymes concentration (serum or plasma) at the beginning and at the end of the trial, as well as the mean changes and SDs of biomarkers' levels.

| Risk of bias assessment
Risk of bias assessment of the included researches was assessed based on the Cochrane criteria (Higgins & Green, 2011). The following items were considered for evaluation of risk of bias of each study: (a) random sequence generation; (b) allocation concealment; (c) blinding of participants and personnel; (d) blinding of outcome assessment; (e) incomplete outcome data; (f) selective outcome reporting; and (g) other potential sources of bias. Based on the Cochrane Handbook recommendations, trials were rated on each the item as "yes" demonstrating low risk of bias, "no" indicating high risk of bias or "unclear" when the risk of bias was unclear or unknown.

| Data synthesis and analysis
Difference in means was defined as effect sizes. Weighted mean differences (WMDs) were calculated as follows: mean divided by the standard deviation of a difference between two random F I G U R E 1 Flow chart of studies selection process Note: All values expressed as mean ± standard deviation (SD) except **for study number 3: Values presented as mean (95% CI).

TA B L E 1 (Continued)
values each from one of two groups (Higgins, 2011). In trials that the standard error (SE) value was reported, SD was obtained using the following formula: SD = SE × √n (n = number of participants in each group). The random-effects model was applied to compute the WMDs with 95% confidence intervals (CIs) for performing the meta-analysis, which took the between-study heterogeneity into account (Borenstein et al., 2009). Heterogeneity of trials was also assessed using Cochran's Q test and I-squared (I 2 ) statistic. Heterogeneity was defined as follow: Q statistic p value of <.1; weak heterogeneity: I 2 = 25-50, relatively high heterogeneity: I 2 = 50-75, high heterogeneity: I 2 = 75-100 (Higgins & Thompson, 2002;Sangsefidi et al., 2020). Subgroup analysis was also accomplished to identify the possible sources of heterogeneity among the selected studies. Since type of intervention (pure ACNs or products rich in ACNs including extracts, beverages, powders, or juices), the administered ACNs dosage, health status of participants (healthy, unhealthy [liver disease, other disease]), trial duration, and assessing liver enzymes as primary or secondary outcomes might have influenced on the results regarding to impact of ACNs, subgroup analysis was carried out based on these variables. Moreover, publication bias was assessed by evaluation of the funnel plot and asymmetry tests including Begg's rank correlation test and Egger's regression test (using p value of <.05) (Duval & Tweedie, 2000). Sensitivity analysis was also conducted to specify the impact of a specific study or a particular group of studies via individual removal of each trial or a specific group and recalculation of the pooled estimates.
Moreover, meta-regression was conducted to evaluate relation of the estimated effect size with ACNs dosage and trial duration.
Statistical analyses were conducted using STATA software, version 11.2 (STATA Corp.). Statistically significant levels were considered as p < .05.

| Study selection and characteristics
Our electronic search of several databases including PubMed, Web of Science, Scopus, and Google scholar resulted in 2,474 articles.
1,375 studies remained after excluding duplicates. Of this numbers, 1,358 researches were excluded since they were not clinical trials (n = 1,328) or did not meet the inclusion criteria (n = 32).
Eventually, 15 surveys met the inclusion criteria and entered in meta-analysis ( Figure 1). We investigated 18 studies in our systematic review as we could not find full text of one article (Kano et al., 2018). For meta-analysis, we entered only 15 studies since we did not achieve detailed data of two researches (Hassellund et al., 2013;Kianbakht & Hashem-Dabaghian, 2019) and exclude one study (Guo et al., 2014) because it affected the results due to crossover design. Characteristics of the included trials are presented in Tables 1-3. All researches were published from 2008 to 2020. The total number of participants in the included trials who completed the surveys was 1,028 (n = 514 in the intervention group, n = 514 in the placebo group). Design of all RCTs was parallel except two studies (Guo et al., 2014;Hassellund et al., 2013) that had cross-over design. Furthermore, most of participants were patients with different diseases such as NAFLD (Chang et al., 2014;Guo et al., 2014;Zhang et al., 2015), pre-diabetes, or diabetes type 2 (Kianbakht et al., 2013;Soltani et al., 2015;Stote et al., 2020;, dyslipidemia (Kianbakht et al., 2014;Qin et al., 2009;Soltani et al., 2014), overweight and obese adults (Wright, Netzel, & Sakzewski, 2013), and pre-hypertension (Hassellund et al., 2013) or hypertension (Kianbakht & Hashem-Dabaghian, 2019;Mohtashami et al., 2019). Nevertheless, three surveys (Curtis et al., 2009;Oki et al., 2016;Suda et al., 2008) had been studied healthy subjects. In addition, only five studies (Chang et al., 2014;Oki et al., 2016;Stote et al., 2020;Suda et al., 2008;Zhang et al., 2015) evaluated liver enzymes as primary outcomes among all researches. Trials durations varied from 28 to 90 days. Dose of administered ACNs was also from 0.77 to 640 mg/day.

| Effect of ACNs on ALT
Meta-analysis of 15 eligible studies (total n = 1,028, intervention:

F I G U R E 2
Forest plot illustrates weighted mean difference (represented by the black square) and 95% confidence interval (CI; represented by horizontal line) for Alanine Aminotransferase (ALT) concentration and anthocyanins. Weights are from random-effects analysis. The area of the black square is proportional to the specific study weight to the overall meta-analysis. The center of the diamond displays the pool weighted mean differences and its width shows the pooled 95% CI TA B L E 5 Subgroup analysis to assess the effect of supplementation with anthocyanins on liver enzymes levels Abbreviations: ALT, alanine aminotransferase; AST, aspartate aminotransferase; CI, confidence interval; GGT, gamma-glutamyl transferase; WMD, weighted mean difference.
*p value ˂.05 considered as significant statistical level. significant association was found between duration of trial and ALT
However, this result was close to the significant level. According to sensitivity analysis, the result of meta-analysis was sensitive to the studies of Kianbakht et al. (2013) and  ( Figure S4). Furthermore, a significant heterogeneity was observed across the studies (p < .0001, I 2 = 91.96.). Subgroup analysis showed that the impact of ACNs did not significantly differ in various doses Q statistics (p) < .0001) participants (Table 5). In addition, ACNs had a significant reducing effect on AST levels in the studies which their primary outcomes were liver enzymes (WMD = −3.464 U/L, 95% CI = −6.034 to 0.894, p = .008) versus the studies which liver enzymes were their secondary outcomes (WMD = −0.577 U/L, 95% CI = −1.957 to 0.802, p = .412; Table 5). Intake of ACNs was also associated with decreased AST levels in the studies which their intervention was products rich in ACNs (WMD = −2.201 U/L, 95% CI = −4.275 to −0.127, p = .037) versus the studies with pure ACNs (WMD = −0.121 U/L, 95% CI = −1.893 to 1.651, p = .8894; Table 5).

F I G U R E 3
Forest plot illustrates weighted mean difference (represented by the black square) and 95% confidence interval (CI; represented by horizontal line) for Aspartate Aminotransferase (AST) concentration and anthocyanins. Weights are from random-effects analysis. The area of the black square is proportional to the specific study weight to the overall meta-analysis. The center of the diamond displays the pool weighted mean differences and its width shows the pooled 95% CI

| Publication bias
According to the funnel plots and asymmetry tests, no significant  Figure S6a). After adjusting the effect size for potential publication bias by the "trim and fill" correc-  Figure S7).

| D ISCUSS I ON
Our meta-analysis revealed no significant effect of consuming ACNs on liver enzymes concentrations. However, intake of ACNs was significantly associated with the reduced levels of ALT and AST in the studies that evaluated liver enzymes as their primary outcomes.
Moreover, ACNs had a significant decreasing effect on AST levels among healthy individuals and in the studies that used ACNs-rich products rich as intervention. We observed no significant association between the dose of ACNs and the levels of liver enzymes.
Significant associations were found between the duration of studies conducted with ALT and AST levels, while no significant relation was detected between the duration and GGT concentrations.
Nevertheless, the findings of ALT should be stated carefully due to the publication bias.
In our study, sensitivity analysis indicated that the results of meta-analysis regarding AST were sensitive to the studies conducted by Kianbakht et al. (2013) and . In the study by Kianbakht et al. (2013), the AST levels increased markedly in the extract rich in ACNs group at the end of trial, although it was not statistically significant. Moreover, the findings of Yang, Ling,  showed a statistically significant increase in the AST levels in ACNs group at the end of study. These findings might be related to the results of sensitivity analysis associated with AST.
Similar to our findings, some meta-analyses showed that ACNs had a significant decreasing effect on fasting blood sugar  represented by horizontal line) for Gamma-Glutamyl Transferase (GGT) concentration and anthocyanins. Weights are from random-effects analysis. The area of the black square is proportional to the specific study weight to the overall meta-analysis. The center of the diamond displays the pool weighted mean differences and its width shows the pooled 95% CI & Shah, 2018;, total cholesterol (Liu et al., 2016;, tumor necrosis factoralpha (TNFα) (Shah & Shah, 2018), and blood pressure (Zhu et al., 2016). Some others reported the increasing impact of ACNs on high-density lipoprotein (Liu et al., 2016;Shah & Shah, 2018).
Moreover, intake of products containing anthocyanins resulted in a significant decrease in ALT levels in NAFLD patients (Zhang et al., 2015), healthy subjects with borderline hepatitis (Oki et al., 2016;Suda et al., 2008), and diabetic patients (Stote et al., 2020) in some RCTs. In the same vein, a significant reduction was found in AST concentrations in healthy individuals with borderline hepatitis (Oki et al., 2016;Suda et al., 2008), and diabetic patients after consumption of products containing anthocyanins.
Moreover, consuming elderberry extract had no significant impact on GGT levels among healthy postmenopausal women (Curtis et al., 2009 In the present meta-analysis, among the included trials, only five studies (Chang et al., 2014;Oki et al., 2016;Stote et al., 2020;Suda et al., 2008;Zhang et al., 2015) investigated liver enzymes as their primary outcomes, while others considered liver enzymes as the secondary outcome. On the other hand, most studies were not specially designed to investigate liver enzyme. In addition, their sample size and duration may not be sufficient for assessing the factors and making concluding. As a result of these issues, some null results may be achieved.
On the other hand, short duration of some trials, various doses used in interventions, and limited number of participants enrolled in the included RCTs can be mentioned to justify the null findings.
Although mechanisms regarding the effect of ACNs on liver enzymes are unclear yet, some possibilities can be presented in this regard. Evidences demonstrated that increase of the serum liver enzymes may reflect damage to the liver cells caused by some factors such as excess deposit of fat in the liver (Ahn et al., 2014), cellular oxidative stress (Suda et al., 2008), inflammation (Suda et al., 2008), and insulin resistance (Ahn et al., 2014;Bonnet et al., 2011;Zhang et al., 2010). In this regard, several animal and cell line studies showed that ACNs could reduce hepatocellular lipid accumulation via suppression of lipogenesis and promotion of lipolysis Guo et al., 2011Guo et al., , 2012Hwang et al., 2011;Jia et al., 2013;Peng et al., 2011;Salamone et al., 2012).
ACNs also decreased hepatocellular oxidative stress by improving antioxidant responses and protection against reactive oxygen spices (Cho et al., 2011;Valenti et al., 2013;Valentová et al., 2007;Zhu et al., 2012). Furthermore, ACNs reduced cellular inflammation by suppressing NF-κB signaling pathways Valenti et al., 2013), and downregulating the inflammatory genes such as TNFα and IL-6 Valenti et al., 2013). It was also found that ACNs improved hepatic and systemic insulin resistance via up-regulation of the glucose transporter 4 in the peripheral tissues, and activation of adenosine monophosphate-activated protein kinase in the peripheral tissues and liver Li et al., 2017;Valenti et al., 2013). Thus, beneficial impacts of ACNs on liver enzymes might be attributed to the protective effects of ACNs against hepatocellular lipid accumulation, oxidative stress, inflammation, and insulin resistance.
Some strengths of our study included application of a powerful search strategy without any linguistic limitations; subgroup analysis according to the health status of participants; type of studies based on liver enzymes' assessment (as primary or secondary outcomes); type of intervention (pure ACNs or products rich in ACNs); dosage of ACNs supplementation; and trial duration. Nevertheless, the current research suffered from some limitations. First, the included RCTs in our meta-analysis had small sample sizes and limited follow-up periods. The assessed trials were also heterogeneous in terms of various factors such as the participants' characteristics, prescription of products as pure ACNs or sources rich in ACNs, dose and duration of supplementation, as well as composition and amount of ACNs in the administered products. Moreover, the possible minor impacts of other compounds cannot be completely ignored in the sources rich in ACNs such as polyphenols, flavonoids, or other phytochemicals.
In conclusion, ACNs had no significant effect on liver enzymes levels. However, consuming ACNs was significantly related to the decrease of ALT and AST levels in the studies that assessed liver enzymes as primary outcomes. Furthermore, ACNs had a significant reducing effect on AST levels among the healthy individuals and in studies with products rich in ACNs as intervention. We found no significant relationship between the dose of ACNs and the levels of liver enzymes. Significant associations were observed between the duration of studies with ALT and AST levels, while no significant relation was discovered between duration of trials and GGT concentrations. However, the findings related to ALT should be reported and applied carefully due to the publication bias. Further well-designed RCTs are required with larger sample sizes and longer follow-ups to confirm these findings.

ACK N OWLED G M ENTS
The authors thank Nutrition and Food Security Center, School of Public Health, Shahid Sadoughi University of Medical Sciences, Yazd, Iran to support this study.

CO N FLI C T O F I NTE R E S T S
The authors declare that there are no competing interests. Formal analysis (lead); Investigation (lead); Methodology (lead);

AUTH O R CO NTR I B UTI
Writing-review & editing (lead).

DATA AVA I L A B I L I T Y S TAT E M E N T
All data generated or analyzed during this study have been included in this published article.