Systematic review and meta-analysis of Coptis chinensis Franch.-containing traditional Chinese medicine as an adjunct therapy to metformin in the treatment of type 2 diabetes mellitus

Background: In China, Coptis chinensis Franch. (Chinese name: Huanglian) prescriptions (HLPs) are prominent hypoglycemic agents used in glycemic control. However, the curative effect of HLPs as adjunctive therapies for type 2 diabetes mellitus (T2DM) has not been evaluated. Based on a systematic review and a meta-analysis, this study was conducted to assess the effects of HLPs combined with metformin as a reinforcing agent for T2DM. Materials and methods: A total of 33 randomized controlled trials (RCTs) reporting on 2,846 cases concerning the use of HLPs in the treatment of T2DM were identified from the China National Knowledge Infrastructure (CNKI), Weipu (VIP), Wanfang, PubMed, Cochrane Library, and EMBASE databases. Primary outcomes included fasting blood glucose (FBG), 2-h postprandial blood glucose (2hPG), glycosylated hemoglobin, type A1c (HbA1c), fasting serum insulin (FINS), and homeostasis model assessment of insulin resistance (HOMA-IR). Secondary outcomes included total cholesterol (TC), triglyceride (TG), low-density lipoprotein cholesterol (LDL-c), high-density lipoprotein cholesterol (HDL-c), and gastrointestinal dysfunction (GD). Continuous data were expressed as mean differences (MDs) with 95% confidence intervals (CIs). The methodological quality of the included RCTs was assessed by Cochrane evidence-based medicine systematic evaluation. Statistical analysis was performed using the Review Manager and Stata software. The required information size and treatment benefits were evaluated by trial sequential analysis (TSA). The quality of evidence was rated using the Grades of Recommendation Assessment, Development, and Evaluation (GRADE) approach. Results: The results revealed that HLPs are beneficial to improve the following: FBG (MD = −1.16%, 95% CI: −1.24 to −1.07), 2hPG (MD = −1.64%, 95% CI: −1.84 to −1.43), HbA1c (MD = −0.78%, 95% CI:−0.96 to −0.60), FINS (MD = −1.94%, 95% CI: −2.68 to −1.20), HOMA-IR (MD = −0.77%, 95% CI: −1.28 to −0.27), TC (MD = −0.70%, 95% CI: −1.00 to −0.39), TG (MD = −0.57%, 95% CI: −0.74 to −0.40), LDL-c (MD = −0.70%, 95% CI: −0.97 to −0.43), and HDL-c (MD = −0.21%, 95% CI: −0.32 to −0.10) for patients with T2DM. The funnel plot, Egger’s test, and trim-and-fill method indicated a moderate publication bias in the results. The TSA showed that the required sample size of HLPs in improving FBG, 2hPG, HbA1c, FINS, HOMA-IR, TC, TG, LDL-c, and HDL-c could sufficiently draw reliable conclusions. GRADE assessment revealed that the quality of the evidence for the effectiveness of HLPs in improving FBG was moderate, but the quality of evidence for 2hPG, HbA1c, FINS, HOMA-IR, TC, TG, LDL-c, and HDL-c was low, and for GD was very low. Conclusion: The systematic review and meta-analysis suggested that HLPs were beneficial for achieving glycemic control. However, HLPs recommended for T2DM patients have yet to be confirmed because of the poor methodological quality of some trials. Therefore, more RCTs with multicenter and double-blind designs are needed to assess the efficacy of HLPs for patients with T2DM.


Introduction
Diabetes mellitus, which seriously endangers human health, is mainly caused by defects in insulin secretion and insulin action and is characterized by disorders of glucose metabolism. (Lin and Sun, 2010). An International Diabetes Federation survey predicted that patients with diabetes mellitus will exceed 645 million by 2045(Carracher et al., 2018. Generally, more than 90% of diabetes mellitus patients have type 2 diabetes mellitus (T2DM). In addition to following diet and lifestyle guidelines, due to the significant hypoglycemic effect of metformin, it is often recommended to intervene with metformin in patients with T2DM (Sharma et al., 2015;Sanchez-Rangel and Inzucchi, 2017). However, due to the certain limitations of metformin in long term use, options from natural products are being searched to meet the need (Sharma and Prajapati, 2017). In recent decades, traditional Chinese medicine (TCM) and its active ingredients have become increasingly popular in Asian countries, and combined with metformin, is widely used as a reinforcing agent in glycemic control (Pang et al., 2018;Tian et al., 2019;Wu et al., 2019). Ancient TCM theories effectively study a disease as a whole and propose that the pathogenesis of diabetes mellitus lies in damp-heat accumulation in the spleen and stomach (Tong et al., 2009). In classic TCM books, Explanation of Materia Medica (Chinese name: Bencaojing Jizhu) and Tang Materia Medica (Chinese name: Tang Bencao) clarified that the prescriptions containing Coptis chinensis Franch. (Chinese name: Huanglian) can effectively alleviate the symptoms of polydipsia, polyphagia, and polyuria (Tong, 2013). Coptis chinensis Franch. As a treatment for diabetes mellitus and related complications, also has a long history in Japan, Korea, Malaysia, Singapore, and India (Li et al., 2013;Sharma et al., 2021). Modern pharmacological investigations have indicated that some ingredients in Coptis chinensis Franch. such as berberine, jatrorrhizine, coptisine, palmatine, epiberbeine, and polysaccharides, exert significant therapeutic effects on multiple targets to improve islet function and regulate glucose metabolism (Fu et al., 2005;Chen et al., 2012;Wang et al., 2019). For example, alkaloids can help alleviate hyperglycemia by promoting glucose uptake (Yang et al., 2014), polysaccharides can produce antidiabetic activity via its antioxidative effect (Jiang et al., 2015), and berberine can improve insulin resistance by inhibiting the expression of tumor necrosis factor-α and free fatty acids .
Recent studies have indicated that Huanglian prescriptions (HLPs) contribute to enhancing insulin sensitivity, stimulating insulin secretion, protecting β-cells, and regulating glycometabolism disorders (Liu et al., 2010;Yu et al., 2012;Li et al., 2019). Therefore, either as monotherapy or adjunct therapy, HLPs are recognized as the most effective TCM antidiabetic prescriptions for T2DM in China. HLPs, such as Dahuang huanglian xiexin (DHHL) decoction, Gegen qinlian (GGQL) decoction, Huanglian ejiao (HLEJ) decoction, Huanglian jiedu (HLJD) decoction, and Huanglian wendan (HLWD) decoction, have been widely used as adjuvant therapies to metformin for glycemic control (Fan et al., 2017;Li et al., 2017;Song et al., 2022;Wang 2020;Zhou et al., 2022). However, to date, there is no large scale clinical evidence on the inhibitory effects of HLPs on T2DM. Also, no published reports can comprehensively evaluate the intervention and side effects of HLPs on glycolipids. Therefore, we included clinical randomized controlled trials (RCTs) for systematic review and meta-analysis to evaluate the effectiveness of HLPs as adjuvant therapies to metformin for patients with T2DM.

Materials and methods
This study followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, obtaining data from published trials.

Search strategies
All articles were searched using medical subject headings terms and free words in the China National Knowledge Infrastructure (CNKI), Wanfang, Weipu (VIP), PubMed, Cochrane Library, and EMBASE databases. The search period for the encompassed articles from the established time to 30 July 2022. Two authors (Xin Zhai and Linlin Pan) independently searched the related articles regardless of type and language. The following terms were used in English databases: ["Type 2 diabetes" or "Type 2 diabetes mellitus" or "T2DM" or "Non insulin dependent diabetes mellitus" or "Impaired fasting glucose" or "Impaired glucose tolerance" or "Xiaoke"] and ["Random allocation" or "Randomized controlled trial" or "Random" or "Randomized" or "Placebo" or "RCT"] and ["Huanglian"or "Coptis chinensis Franch." or "Coptidis Rhizoma" or "Coptis chinensis" or "Rhizoma coptidis"]. The following terms were used in Chinese databases: ["Erxing Tangniaobing" or "Xiaoke" (T2DM) ] and ["Suiji duizhao shiyan" or "Mangfa" or "Anweiji" (RCT) ] and ["Huanglian"]. The search strategies are presented in detail in Supplementary Table S1.

Literature selection and data extraction
Two authors (Linlin Pan and Xin Zhai) independently evaluated the title, abstract, and full texts of the articles. The articles that met the inclusion criteria were then selected. Inconsistencies were settled by discussion. Finally, important information from the included articles was extracted for analysis, including the name of the first author, year of publication, trial types, sample size, sex, age, course of the disease, interventions, and course of treatment.

Risk of bias
Linlin Pan and Xin Zhai independently evaluated the methodological quality of each trial by using the Cochrane riskof-bias tool (Higgins et al., 2011). Disagreements were discussed and resolved by Guirong Liu. The criteria assessed were as follows: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other biases. The risk of bias was rated as high, unclear, or low.

Data synthesis and analysis
RevMan (version 5.3) was used to perform statistical analysis. Continuous data were expressed as the mean difference (MD) with a 95% confidence interval (CI), and p < 0.05 was considered statistically significant. Heterogeneity was evaluated using the chi 2 and I 2 tests, and p < 0.10% or I 2 > 50% was considered to have marked heterogeneity. The low-heterogeneity data (p > 0.10% or I 2 < 50%) used the fixed-effect model, and the high-heterogeneity data (p < 0.10% or I 2 > 50%) used the random-effects model. Sensitivity analysis was evaluated using various statistical methods. Publication bias was assessed by visual observation of the symmetry of funnel plots, Egger's test (p < 0.05 indicates publication bias), and the trim-and-fill method.
Trial sequential analysis (TSA) was conducted to calculate the required information size (RIS) for meta-analysis and evaluate the intervention benefits on the basis of the accrued information size (AIS). The risk of a type I error was set at 5% with a power of 80%. The variance was calculated based on the data included in the trials, and the relative risk reduction was set at 20% (Wetterslev et al., 2017). The evidence for the intervention was considered reliable when cumulative Z-curves crossed sequential monitoring boundaries. The Grades of Recommendation Assessment, Development, and Evaluation (GRADE) approach was used to rate the quality of the evidence as high, moderate, low, or very low (Guyatt et al., 2008).

Study characteristics
A total of 33 RCTs published from 2006 to 2022 were included in this study. The RCTs consisted of 2,846 patients with T2DM between 18 and 86 years of age (Table 1). All trials were single-center trials, and the detection time ranged from 2 to 24 weeks. A total of 1,437 patients in the experimental group underwent treatment using HLPs plus metformin, and 1,409 patients in the control group underwent metformin treatment. Among the 33 trials, three trials with 268 patients used DHHL decoction, seven trials with 551 patients used GGQL decoction, five trials with 438 patients used HLEJ decoction, eight trials with 885 patients used HLJD decoction, and ten trials with 704 patients used HLWD decoction (Table 2).

Quality assessment
A total of 33 RCTs were identified in this study (Figure 2), of which 16 used the random number table method to generate random sequences (Chen et al., 2012;Zou and Lao, 2016;Fu, 2017;Liu, 2017;Zhang et al., 2018;Feng, 2019;Jin et al., 2019;Wu et al., 2019;Xiong, 2019;Wang 2020;Wang, 2020;Fu, 2021;Liu et al., 2021;Song et al., 2022;Wang et al., 2021), and others only mentioned randomly assigned participants. Three trials used the double-blind method for participants and personnel (Zou and Lao, 2016;Pang et al., 2018;Song et al., 2022), and others provided no detailed information. The risk of detection bias was low in all trials, because FBG, 2hPG, HbA1c, FINS, HOMA-IR, TC, TG, LDL-c, HDL-c, LDL-c, and GD levels were evaluated based on objective criteria. In the study  (2017), three patients in the experimental group and the control group withdrew from the trial (9% exit rate). The remaining trials without the loss of follow-up patients or with the loss of follow-up rate <5% were described as having a low-attrition bias. For the reporting bias, nine trials with only positive results were determined as unclear (Zhang, 2014;Li et al., 2017;Liu, 2006;Liu, 2017;Pang et al., 2018;Zhang et al., 2018;Xiong, 2019;Zhang, 2019). For other bias, ten trials were unclear in the sex of the patient, course of the disease, and course of treatment  Frontiers in Pharmacology frontiersin.org Wang, 2013;Zhang, 2014;Zhang, 2019). Meanwhile, others with detailed information presented a low risk. TSA results revealed that the AIS exceeded the RIS for the effectiveness of HLPs in improving FBG (AIS 2,846 was larger than RIS 268), 2hPG (AIS 2,280 was larger than RIS 338), HbA1c (AIS 2,698 was larger than RIS 834), FINS (AIS 1,191 was larger than RIS 748), HOMA-IR (AIS 943 was larger than RIS 680), TC (AIS 1,235 was larger than RIS 695), TG (AIS 1,127 was larger than RIS 506), LDL-c (AIS 1,235 was larger than RIS 641), and HDL-c (AIS 753 was larger than RIS 204), and their cumulative Z-curves crossed the trial sequential monitoring boundary ( Figures 3A-I), indicating that their current evidence was sufficient to draw a reliable conclusion. However, the AIS didn't exceed the RIS for the effectiveness of HLPs in improving GD ( Figure 3J), indicating that the current evidence was't sufficient to draw a reliable conclusion. GRADE assessment suggested that the quality of evidence was moderate for the effectiveness of HLPs in improving FBG, but the quality of evidence was low for 2hPG, HbA1c, FINS, HOMA-IR, TC, TG, LDL-c, and HDL-c, even very low for GD (Table 3).

HLPs for blood lipids
A total of 16 trials comprising 628 subjects in the experimental group and 607 subjects in the control group evaluated the effectiveness of HLPs in improving TC ( Figure 4F). Subgroups were divided depending on the type of HLPs for TC. The results revealed that patients who received metformin in combination with GGQLdecoction (MD = −0.57%, 95% CI:−0.99 to −0.15, and p = 0.008), HLEJ decoction (MD = −1.38%, 95% CI:−1.62 to −1.14, and p < 0.00001), and HLJD decoction (MD = −1.53%, 95% CI:−1.87 to −1.19, and p < 0.00001) respectively were more likely to have reduced TC relative to those with metformin alone. No significant heterogeneity was indicated in HLJD decoction for TC (I 2 = FIGURE 4 (Continued).

HLPs for GD
A total of seven trials comprising 313 subjects in the experimental group and 294 subjects in the control group conducted analysis of HLPs for GD ( Figure 4J). Patients who received HLPs can't reduce GD relative to those with metformin alone (OR = 0.54%, 95% CI:0.26 to 1.10, and p = 0.09).

Sensitivity analysis
The results in Table 4 suggest that patients with T2DM in the experimental group show improved FBG, 2hPG, HbA1c, FINS, HOMA-IR, TC, TG, LDL-c, and HDL-c relative to those in the control group. However, changes in the effectiveness of HLPs in improving 2hPG, HbA1c, FINS, HOMA-IR, TC, TG, LDL-c, and HDL-c showed significant heterogeneity. With regard to the subgroup sensitivity analysis, after excluding some underestimated or overestimated trials, the heterogeneity of the majority of studies was significantly reduced, including the following: HLJD for 2hPG, HLWD for FINS; GGQL, HLEJ, HLJD, and HLWD for HbA1c; GGQL and HLEJ for TC; GGQL and HLEJ for TG; HLEJ and HLWD for LDL-c; and DHHL for HDL-c. However, no statistically significant difference was found in DHHL for 2hPG, HLJD for FINS, and DHHL for HOMA-IR.

Publication bias
As shown in Figure 5, the funnel plots used to evaluate the effectiveness of HLPs in improving FBG are nearly Frontiers in Pharmacology frontiersin.org symmetrical, whereas those used to assess the effects of HLPs on 2hPG, HbA1c, FINS, HOMA-IR, TC, TG, LDL-c, and HDL-c are asymmetrical. Therefore, Egger's test (Stata version 13.0) was also performed to evaluate their publication bias. The Egger's test used to assess publication bias suggested that p > 0.05 in FBG, 2hPG, TC, TG, LDL-c, and HDL-c, whereas p < 0.05 in HbA1c, FINS, and HOMA-IR (Table 5). Finally, the trim-and fillmethod (Stata version 13.0) was used to evaluate the publication bias of HbA1c and HOMA-IR. In Figure 6A

Discussion
The potential of HLPs to prevent and treat T2DM has been investigated in several studies, and its hypoglycemic mechanism is becoming increasingly apparent. DHHL decoction can regulate the glucose level by activating AMPKα and upregulating the expression of PGC-1α and GLUT4 (Hao et al., 2019). GGQL decoction can enhance glucose metabolism by regulating tryptophan, pantothenic acid, and adenine in IR-HepG2 cells , as well as improve liver insulin resistance by upregulating SIRT1 expression and reducing FoxO1 acetylation . HLEJ decoction can exert glucose-lowering and lipid-lowering effects by resisting inflammation and improving insulin resistance (Feng, 2015). HLJD decoction can exert hypolipidemic effects by inhibiting the increased activity of intestinal pancreatic lipase (Zhang et al., 2013) and increasing GLUT4 and PI3K p85 mRNA expression in adipose and skeletal muscle tissues Jin et al., 2007). HLWD decoction can effectively treat glycometabolism disorder by repairing the insulin signaling pathway and inhibiting the release of inflammatory cytokines (Li et al., 2016;Chen et al., 2019).  Frontiers in Pharmacology frontiersin.org 18 In this study, treatment with different HLPs exhibited different hypoglycemic and lipid-lowering effects, suggesting that metformin combined with different HLPs may cause variations in medicinal metabolism. This study found that DHHL decoction can improve FBG, 2hPG, HbA1c, and FINS, but does not affect HOMA-IR. In addition, no well-established data are available to analyze the effect of DHHL decoction on TC, TG, LDL-c, and HDL-c. GGQL decoction can improve all blood glucose and blood lipid indicators. HLEJ decoction can improve FBG, 2hPG, HbA1c, TC, TG, and LDL-c, but its role in FINS, HOMA-IR, and HDL-c has not been reported. HLJD decoction can improve FBG, 2hPG, HbA1c, HOMA-IR, TC, TG, LDL-c, and HDL-c, but exerts no effect on FINS. HLWD decoction can improve FBG, 2hPG, HbA1c, FINS, HOMA-IR, and TG, but the  improvement in TC, LDL-c, and HDL-c was not statistically significant. Therefore, among all HLPs, GGQL decoction is potentially the most effective prescription for improving T2DM.
The advantages of this study are as follows: 1) In the sensitivity analysis, the difference in prescriptions may be the important source of heterogeneity, so we performed a subgroup analysis in different HLPs. Meanwhile, the overall results exhibited heterogeneity in this study, so we excluded the individual trials that caused heterogeneity, and the heterogeneity was significantly reduced. 2) With regard to publication bias, we used funnel plot, Egger's test, and trim-and-fill method to evaluate the publication bias. The results of funnel plot and Egger's test suggest that no publication bias was found in the enhancing effect of HLPs on FBG, 2hPG, TC, TG, LDL-c, and HDL-c. Then the trim-and-fill method was used to further evaluate the publication bias of HbA1c, FINS, and HOMA-IR, which still has important reference significance for the improvement of HbA1c, FINS, and HOMA-IR with HLPs. 3) This study also applied TSA analysis to assess the sample size required and thereby draw reliable conclusions. The sample size of all but one (HLPs for GD) were found sufficient to support this study and thereby draw reliable conclusions. Therefore, the results of this study present high reliability.
The present study also has several limitations: 1) All RCTs included in this study were Chinese, which likely led to geographical bias. Thus, an international collaboration should be conducted to ensure the generalizability of the findings. 2) The methodological quality of the RCTs was low, only half of the RCTs described the allocation concealment and blinding method, which might have led to a nonnegligible risk of bias. Thus, more scientific RCTs with specific randomize allocation details are needed. 3) Different kinds of HLPs vary in their hypoglycemic mechanism of action. Thus, high heterogeneity was observed among different HLPs, limiting the confirmation of the efficacy of HLPs in the treatment of T2DM. 4) Variations in dose in the same prescription are a concern in TCM. Variations in dose may also lead to differences in efficacy, leading to heterogeneity in research. 5) Current evidence shows that GGQL decoction can be potentially used as the optimal complementary approach to regulate glucose and lipid levels, but this finding has yet to be proved. Therefore, more rigorously designed and large-scale RCTs are required to confirm our findings.

Conclusion
Current evidence from this meta-analysis and systematic review suggests that compared with metformin alone, HLPs provide more benefits for the treatment of T2DM, particularly in FBG, 2hPG, HAb1c, FINS, HOMA-IR, TC, TG, LDL-c, and HDL-c. Due to insufficient data from the included RCTs, the therapeutic effect of HLPs on GD has not been demonstrated, and the findings should be elucidated with caution because of the limitations. Therefore, larger-scale and well-designed RCTs are essential to verify HLPs as a promising candidate treatment for patients with T2DM.

FIGURE 6
The trim and fill analysis of HbA1c and HOMA-IR. Note: (A) Represents the trim and fill analysis of HbA1c; (B) Represents the trim and fill analysis of FINS; (C) Represents the trim and fill analysis of HOMA-IR. The circle represents the actual estimate and the square represents the theoretical estimate when publication bias does not exist.
Frontiers in Pharmacology frontiersin.org