We retrieved from PubMed, EMBASE, Cochrane Library, Web of Science, Scopus, Wan Fang Database, China Science and Technology Journal Database (VIP), China National Knowledge Infrastructure (CNKI), and China Medical Biological Literature Database (CMB) from the time of inception to 20 March 2023. In addition, we scoured ongoing studies on the World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP), ClinicalTrials, and the China Clinical Trials Registry (CHiCTR). The search terms mainly included “gegen qinlian”, “Gegen Qinlian decoction”, “Gegen Qinlian tang”, “Type 2 Diabetes Mellitus”, “Type 2 Diabetes”, and “Diabetes Mellitus, Non-Insulin Dependent”. The detailed search strategy for the search terms is described in Supplementary Material .

The inclusion criteria were as follows. (1) For the study design, all published RCTs of GQD or modified GQD for the treatment of patients with T2DM were included. Publication language was limited to English or Chinese. (2) The study objects were individuals with a diagnosis of T2DM in adults (18 years of age or older). (3) For the study intervention, GQD or modified GQD was used in the treatment group in any dosage form and amount. The control group was provided with placebo or the conventional treatment, including health education, dietary management, exercise intervention, blood glucose monitoring, and hypoglycemic medication. The treatment group may also use interventions from the control group, but must be consistent with the control group. (4) For the study outcomes, primary outcomes were fasting blood glucose (FBG), 2-h postprandial glucose (2hPG), and glycosylated hemoglobin (HbA1c); secondary outcomes were total cholesterol (TC), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), and homeostasis model assessment of insulin resistance (HOMA-IR). Safety outcomes were any adverse events.

The exclusion criteria were as follows: (1) non-RCTs, conference abstracts, cohort studies, cross-sectional studies, case reports, guidelines, animal studies, and review articles; (2) other types of diabetes mellitus and patients with acute metabolic disorders, severe hepatic or renal impairment, severe cardiovascular disease, pregnancy, or lactation; (3) as to intervention, exclude studies using traditional Chinese medicine (TCM) treatments other than GQD; (4) for outcomes, those with incorrect data and incomplete measurement of results; and (5) repeated articles.

A database was created using EndNote X20 to manage and filter database records. Data were extracted by two authors (Yi-Mei Tan and Shuang-Hua Liu), and any inconsistencies were resolved after a debate with a third investigator (Qi-Zhi Tang). Data extraction items included first author, year of publication, study design, diagnostic criteria, sample size, sex, average age, duration of illness, duration of treatment, interventions, outcomes, comorbidities, and adverse events.

Two authors (Yi-Mei Tan and Shuang-Hua Liu) independently assessed the risk of bias using the Risk of Bias Assessment Tool from the Cochrane Handbook. Disagreement was resolved through discussion with another reviewer (Qi-Zhi Tang). The risk of bias was assessed through seven dimensions: (1) random sequence generation, (2) allocation concealment, (3) blinding of participants and personnel, (4) blinding of outcome assessments, (5) incomplete outcome data, (6) selective reporting, and (7) other biases.

Meta-analysis was performed using RevMan 5. 3 and Stata 14.0. Relative risk ratios (RR) were used for dichotomous variables. For continuous variables, mean difference (MD) was used when the units of the outcome indicator were the same; otherwise, standardized mean difference (SMD) was used, and 95% confidence intervals (CI) were given. The chi² test and I² test were used to test for heterogeneity among studies. If p ≤ 0.05 and I²≥ 50%, this indicates statistically significant heterogeneity between studies; thus, a random effects model was used. We planned to explore the sources of heterogeneity and judge the stability by performing subgroup analyses and sensitivity analyses. We performed meta-regressions on the outcome metrics including more than 10 studies (FBG, 2hPG, and HbA1c) in terms of sample size, year of publication, and average age. In addition, we performed funnel plot and Egger’s linear regression tests for publication bias for FBG, 2hPG, and HbA1c, and p<0.05 was considered statistically significant, indicating possible publication bias. We conducted subgroup analyses based on the following predefined subgroup hypotheses: (1) average age (≤60 years or >60 years), (2) duration of T2DM (≤5 years or >5 years), and (3) duration of treatment (≤2 months or >2 months).

A total of 1,455 studies were retrieved through a database search. Three original texts were not available because they were missing from the database export. After removing 832 duplicates, 507 studies were excluded after checking the titles and abstracts of 620 citations. The full text of the remaining 113 studies was read, and 96 studies were excluded based on inclusion and exclusion criteria. Ultimately, 17 eligible studies were included in the quantitative analysis. Details of the literature screening procedure are shown in Figure 1 .

A total of 17 studies including 1,476 patients with T2DM were included, 830 men and 646 women (30–46). All studies were conducted in China and spanned from 2012 to 2022. The average age of the participants ranged from 36.4 to 71.9 years old, the duration of the disease ranged from 0.26 to 12.5 years, and the treatment period varied from 3 weeks to 3 months. There were nine studies with treatment groups using the original GQD; the others used modified GQD. The composition of GQD or modified GQD is shown in Supplementary Material . There were 13 studies where the treatment group was treated with GQD or modified GQD combined with conventional treatment. While there were three studies that the treatment group was treated with GQD or modified GQD. Only one study adopted a placebo-controlled clinical trial design approach. GQD or modified GQD is prescribed as one dose once or twice a day. Doses of s range from 9 g to 60 g, Scutellaria baicalensis and Rhizoma Coptidis from 6 g to 22.5 g, and licorice from 4 g to 15 g. The basic characteristics of the included studies are shown in Table 1 .

Three studies did not report the specific methodology used in the generation of the randomized sequences (33–35). Randomization method of allocation concealment was described in three studies (32, 40, 45). One study used a stratified randomized, double-blind, placebo-controlled, multicenter clinical trial design methodology (45). One study did not report the number of patients at the end of the study (35). None of the studies described the blinding of outcome assessment, selective reporting. Overall, the methodological quality of the included literature was suboptimal. The results of the risk of bias assessment of the included studies are shown in Figure 2 .

Adverse events were reported in 10 of the 17 included studies. The results of two studies showed that adverse events were significantly lower in the combination group than in the conventional treatment group (44, 46). The results of other studies showed that the incidence of adverse events with GQD alone or in combination with conventional treatment was not significantly different from that of the conventional treatment group or the placebo group. A summary table of adverse events is available in Supplementary Material . The results of the meta-analysis of adverse events suggest that GQD is relatively safe ( Figure 8 ).

Funnel plots and Egger’s test were used to assess publication bias for FBG, 2hPG, and HbA1c ( Figure 9 ). Funnel plots for HbA1c showed all but one study clustered on the top of the funnel plots, and two studies deviating from the pooled effect sizes, suggesting that there may have been heterogeneity among the studies. The Egger’s test showed no statistical difference (p=0.209), indicating that there was no significant publication bias in the studies of HbA1c. The funnel plots of FBG and 2hPG showed roughly symmetrical distributions, consistent with the results of Egger’s test (p=0.153 and 0.851, respectively), suggesting that there was no significant publication bias in the studies of FBG and 2hPG.

T2DM is a group of metabolic diseases characterized by chronic hyperglycemia caused by multiple etiologies, which is due to defective insulin secretion and/or utilization (47, 48). Lots of studies have found that GQD has the efficacy of regulating glycolipid metabolism and improving clinical symptoms (47, 49).

In this study, we included 17 studies and analyzed the efficacy and safety of GQD in the treatment of T2DM. Our main finding was that GQD combined with conventional treatment could reduce FBG, 2hPG, and HbA1c. GQD alone has an adjunctive hypoglycemic effect, but the quantity and quality of the included literature is small, and more high-quality studies are needed to confirm this. We suppose GQD combined with conventional treatment can be a useful complementary treatment for T2DM. There was a large heterogeneity in the results of FBG and HbA1c in the GQD combined with conventional treatment. We did not find sources of heterogeneity through meta-regression and subgroup analysis. BMI, comorbidities, and islet function are likely sources of heterogeneity. In addition, methodological shortcomings of the included studies, such as lack of blinding and allocation concealment, may have contributed to heterogeneity.

In terms of lipid metabolism, combination treatment was superior to conventional treatment in reducing TC, TG, and LDL-C and improving HDL-C. This suggests that patients with T2DM combined with abnormal lipid metabolism can benefit not only in terms of lowering blood glucose but also improving lipid metabolism with GQD combination treatment. GQD alone had some adjunctive improvement in lipid metabolism, but due to the limited number of included studies, we do not know whether GQD alone has an improved effect on lipid metabolism.

In addition, GQD combined with conventional treatment was better than conventional treatment in reducing HOMA-IR. HOMA-IR is highly heterogeneous; subgroup analyses did not identify sources of heterogeneity. We hypothesize that the heterogeneity may be related to individual differences, or it may be measurement bias caused by differences in insulin assay methods. In addition, the quality of the studies that we included was low, possibly because of the lack of blinding and allocation concealment leading to heterogeneity. The available research suggests that GQD improves islet function and controls blood glucose levels in T2DM patients (50, 51). But due to the small number of studies, it is not clear whether GQD alone has an improving effect on HOMA-IR, and more studies are needed to demonstrate this.

Adverse events were assessed in 10 of the 17 included studies. All of the adverse reactions reported in the studies resolved on their own, and no specific treatment was given. No serious adverse events were observed. The meta-analysis results suggest that GQD is relatively safe if used correctly. We performed funnel plots and Egger’s test for FBG, 2hPG and HbA1c, and no publication bias was found, suggesting that the results have some reliability.

GQD can improve insulin resistance by activating the expression of GPR119, promoting the secretion of intestinal GLP-1 (28), increasing the serum superoxide dismutase (SOD) content, and decreasing the level of malondialdehyde (MDA) (14), and improving the inflammatory factors, such as PS, TNF-α, and IL-6, and regulating the structure of intestinal flora (16, 25, 27, 52) Through network pharmacology and bioinformatics analysis, GQD modulated 82 diabetes-related proteins and 59 diabetes-related biological pathways. Among them, modulation of the ESR1 signaling pathway plays an important role in the mechanism of GQD treatment of T2DM (53). Zhou et al. suggested that GQD may protect pancreatic islet β-cells by increasing the activity of the IRS-2/PI3K-Akt signaling pathway (54).