Efficacy of Xiaokeqing granules and lifestyle intervention in treating prediabetes mellitus considering metabolomic biomarkers: A randomised controlled trial

Abstract

BACKGROUND

Prediabetes mellitus (PDM) is receiving increasing attention as a precursor to type 2 diabetes mellitus. Lifestyle and traditional Chinese medicine (TCM) interventions are effective for PDM prevention and treatment. Therefore, we conducted a preliminary investigation and an exploratory randomised controlled trial to assess the effects of a combined lifestyle and TCM intervention on PDM indicators.

AIM

To study the effectiveness of Xiaokeqing granules (XQG) and lifestyle interventions in PDM participants while using metabolomics to identify potential markers.

METHODS

Forty PDM participants with yin deficiency syndrome with excessive heat were recruited and randomly allocated to the control (Con) group or the XQG group (20 per group). The Con group underwent lifestyle interventions, whereas the XQG group underwent lifestyle and XQG interventions. The follow-up duration was 2 months. Fasting blood glucose, 2-hour postprandial glucose (2hPG), glycated haemoglobin A1c, fasting insulin, homeostasis model assessment-insulin resistance levels, and serum metabolomics characteristics were compared via liquid chromatography-tandem mass spectrometry analysis.

RESULTS

There were significant differences in 2hPG between the two groups (P < 0.05) in the intention-to-treat analysis and per-protocol analysis. The intervention method used in this study was safe (P > 0.05). Groenlandicine, kaempferol, isomangiferin, etc., are the XQG constituents absorbed in the blood. N-Nervonoyl methionine and 5-hydroxy-L-tryptophan are core potential metabolomic biomarkers for the effectiveness of XQG and lifestyle interventions. HTR1A, HTR2C, SLC6A4, etc., are the core targets of XQG and lifestyle interventions, as well as the reason for their clinical efficacy. Possible mechanistic pathways include tryptophan metabolism, pantothenate and certificate of analysis biosynthesis, lysine degradation and biosynthesis of cofactors.

CONCLUSION

This pilot study provides evidence that a combined XQG and lifestyle intervention can improve 2hPG in participants with PDM. The mechanism of action is related to multiple constituents, targets and pathways.

Key Words: Prediabetes mellitus; Xiaokeqing granules; Lifestyle; Two-hour postprandial glucose; Metabolomics; Tryptophan metabolism

Core Tip: In this study, Xiaokeqing granules constitute a new class-three traditional Chinese medicine. They can improve 2-hour postprandial glucose in participants with prediabetes mellitus with yin deficiency syndrome with excessive heat according to a preliminary investigation and an exploratory randomised controlled trial. Groenlandicine and kaempferol are the main Xiaokeqing granule constituents absorbed into the blood. N-Nervonoyl methionine and 5-hydroxy-L-tryptophan are potential biomarkers for the effectiveness of Xiaokeqing granules and lifestyle interventions.

INTRODUCTION

With the rapid increase in the incidence and prevalence rates of type 2 diabetes mellitus (T2DM) in China, prediabetes mellitus (PDM) has attracted increasing attention[1]. An investigation revealed that the prevalence rate of PDM is 15.5%-50.1%[2,3]. PDM is characterized by impaired fasting glucose, glucose tolerance, and glucose regulation. The PDM population is the primary source of patients with T2DM. Every year, approximately 10% of PDM population progress to T2DM, which results in heavy medical and social burdens. Therefore, research on screening and prevention in the PDM population is highly important[4-7]. A series of studies have shown that interventions for population with PDM involving lifestyle or drugs can delay entry into the T2DM stage or reverse PDM[8-12]. Long-term adherence to diet and exercise interventions or the use of hypoglycaemic drugs can involve risks, including reduced quality of life, hepatotoxicity, nephrotoxicity, gastrointestinal adverse reactions, etc., resulting in poor compliance; thus, population with PDM cannot adhere to these interventions for a long period, which affects long-term benefits.

PDM is classified into the “spleen dan” category in traditional Chinese medicine (TCM). The use of TCM for interventions in population with PDM conforms to the idea of “preventing disease”. A series of studies have shown that TCM can prevent and treat PDM, mainly by reducing fasting blood glucose (FBG), 2-hour postprandial glucose (2hPG), and glycated haemoglobin A1c (HbA1c) levels and improving clinical symptoms[13-15]. Xiaokeqing granule (XQG) is a new class-three TCM drug (national medical product administration approval number: Z20080034) developed by Tasly Pharmaceutical Group Co., Ltd. (Tianjin, China). It has the effects of nourishing yin and clearing heat. XQG include Anemarrhena asphodeloides Bunge, Atractylodes lancea (Thunb.) DC, Coptis chinensis Franch, Typha angustifolia L, and Euphorbia humifusa Willd. The ratio is 10:6:1:6:6. Previous research has shown that XQG have a hypoglycaemic effect[16,17]. However, the role and mechanism of XQG in treating individuals with PDM are still unclear. Therefore, in this study, we investigated the effects of XQG and lifestyle interventions on blood glucose in PDM participants through a small-sample, randomised controlled trial and objectively evaluated the clinical efficacy and safety of XQG. Moreover, via nontargeted metabolomics, the effects of XQG and lifestyle interventions on serum metabolites were observed to explore the possible mechanism involved in improving blood glucose.

MATERIALS AND METHODS

Participants

The participants were recruited by clinical researchers from the Department of Endocrinology of the First Affiliated Hospital of Anhui Chinese Medicine University between July 1, 2023, and June 30, 2024. The inclusion criteria were as follows: Diagnosed with PDM and met at least two of the three following conditions: 6.1 mmol/L ≤ FBG < 7.0 mmol/L, 7.8 mmol/L ≤ 2hPG < 11.1 mmol/L, and 5.7% ≤ HbA1c < 6.5%[18]. The TCM syndrome type is yin deficiency syndrome with excessive heat, including thirst, drinking a lot, eating a lot, being easily hungry, fear of heat, irritability, constipation, and so on[16]. The age range was 18-70 years, and sex was not limited. The participants voluntarily signed the informed consent form and indicated their willingness to comply with the intervention plan and cooperate with the follow-up person.

The exclusion criteria were as follows: Hepatic failure [alanine aminotransferase (ALT) and aspartate aminotransferase (AST) > 3 times the upper limit of normal], renal failure [blood urea nitrogen (BUN) and creatinine (Cr) > 2 times the upper limit of normal], acute cardiovascular and cerebrovascular events or myocardial infarction, a state of stress or secondary hyperglycaemia, mental illness, pregnancy or lactation, plans to conceive or no contraception plan, potential allergy to XQG, and other serious primary diseases.

The exclusion and dropout criteria were as follows: (1) Participants who did not use the XQG or lifestyle intervention after randomization; (2) Participants with no post-treatment visit data or who were unable to attend follow-up; and (3) Participants who violated the efficacy evaluation and safety judgment protocols.

Study design

This study is a prospective, preliminary investigation and an exploratory randomised controlled trial. The participants were allocated to the control (Con) and XQG groups by the simple random assignment method, performed by data analysis professionals. These professionals used a computer program to generate a random number of 0 or 1 for each participant, with a final 1:1 ratio. Clinical researchers assigned the participants to the interventions. This study followed the Declaration of Helsinki and the Guidelines for Good Clinical Practice. The ethical approval number 2022AH-73 was given by the ethics committee of the First Affiliated Hospital of Anhui Chinese Medicine University. The registration number ITMCTR2023000057 was approved by the International Traditional Medicine Clinical Trial Registration Platform.

Sample size

We calculated 18 participants for each group, with a test family: F tests. Statistical test: Repeated-measures analysis of variance between factors. Type of power analysis: A priori calculation of the required sample size given α = 0.05, power = 0.85, effect size f = 0.3, number of groups = 2, and number of measurements = 3. This calculation method is based on GPower software (Heinrich-Heine-Universität Düsseldorf, Germany). Considering a sample loss of 10%, the number of participants needed for each group was calculated to be 20. Finally, we recruited 40 participants for the Con and XQG groups[19].

Intervention

Clinical researchers assigned participants to interventions. The Con group interventions were conducted according to the lifestyle of health management in the PDM guidelines, which mainly include becoming familiar with the basic knowledge of PDM medical nutrition, reasonably controlling energy intake, and evenly distributing various nutrients throughout the day. In terms of exercise, participants were instructed to perform at least 150-300 minutes of moderate-intensity aerobic exercise and 2 resistance exercise sessions per week and to reduce their sitting time[20,21]. The XQG group underwent the same lifestyle intervention as the Con group but were also administered oral XQG (Tasly Pharmaceutical Group Co., Ltd., China), 6 g/time, three times/day. The preparation method and processing of XQG were in accordance with the national standard of the National Medical Products Administration (YBZ00432008). Moreover, the ultra-performance liquid chromatography-photo-diode array-evaporative light-scattering detector fingerprint spectrum was used for quality control to ensure reproducibility. Follow-up 1 and follow-up 2 were conducted once every month, for a total of 2 follow-ups.

Observation indications

The baseline characteristics included age, sex, body mass index (BMI), disease duration, systolic blood pressure (SBP), and diastolic blood pressure (DBP). The glycometabolic variables included FBG, 2hPG, HbA1c, fasting insulin (FINS), and homeostasis model assessment-insulin resistance (HOMA-IR). The safety variables included ALT, AST, BUN, and Cr levels and adverse events. The three categories of variables were measured at baseline and at follow-up 2. FBG and 2hPG were measured at follow-up 1.

FBG, 2hPG ALT, AST, BUN and Cr were measured using a Beckman AU5800 fully automatic biochemical analyser (Beckman, United States). The HbA1c level was determined using a Bio-Rad D-100 HbA1c analyser (Bio-Rad, United States). The FINS was determined using an AutoLumo A2000 Plus fully automatic chemiluminescence analyser (Autobio, Zhengzhou, Henan Province, China)[22]. HOMA-IR = FINS (μIU/mL) × FBG (mmol/L)/22.5.

XQG and serum nontargeted metabolomics analysis

Venous blood samples from the XQG group were collected using biochemical tubes after a 12-hour overnight fast at baseline or after oral administration of XQG for 2 hours at follow-up 1. The blood samples were centrifuged at 4000 revolutions per minute for 10 minutes. Serum was extracted and stored at -80 °C for nontargeted metabolomics analysis.

XQG samples: Fifty milligrams of XQG were added to a 2 mL centrifuge tube, and a 6 mm diameter grinding bead was added. A total of 400 μL of solution was used for metabolite extraction. XQG were ground with a Wonbio-96c frozen tissue grinder (Shanghai Wanbo Biotechnology Co., Ltd.) for 6 minutes at -10 °C and 50 Hz, followed by low-temperature ultrasonic extraction for 30 minutes at 5 °C and 40 kHz. The XQG were left at -20 °C for 30 minutes and then centrifuged for 15 minutes at 4 °C at 13000 × g, after which the XQG supernatant was subjected to liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis. The purpose is to obtain the components of the XQG.

Serum samples: A 100 μL of serum from the XQG group at baseline or after oral administration of XQG at follow-up 1 was added separately to a 1.5 mL centrifuge tube with 400 μL of solution to extract metabolites. The serum samples were mixed by vortexing for 30 seconds and then sonicated at a low temperature for 30 minutes at 5 °C and 40 kHz. The samples were incubated at -20 °C for 30 minutes to precipitate the proteins. The serum was subsequently centrifuged for 15 minutes at 4 °C and 13000 × g, after which the supernatant was blown dry under nitrogen. The sample was then resolubilised with 100 μL of solution and extracted by low-temperature ultrasonication for 5 minutes at 5 °C and 40 kHz, followed by centrifugation at 13000 × g and 4 °C for 10 minutes. The serum supernatant was used for LC-MS/MS analysis. The purpose was to identify which components of the XQG were in the serum of the participants.

The pooled quality control samples were prepared by mixing equal volumes of all samples from the same period. The purpose is to evaluate the stability of the detection method. Internal standards were used in the research system to evaluate the stability of the experimental process. The pretreatment of the raw LC-MS/MS data was performed by Progenesis QI (Waters Corporation, Milford, United States) software. The three-dimensional data matrix in comma-separated values format information included the metabolite name and mass spectral response intensity. The metabolites were identified by the HMDB, METLIN and Majorbio databases. The data matrix was preprocessed and filtered. Each metabolic signature was normalised to the sum for data analysis.

LC-MS/MS analysis of the samples was conducted on a Thermo UHPLC-Q Exactive system from Majorbio Bio Pharm Technology Co. Ltd. (Shanghai, China). This system is calibrated regularly to ensure data stability.

The analytical LC conditions were as follows: 3 μL of the serum or XQG supernatant was separated on an ACQUITY HSS T3 column (100 mm × 2.1 mm i.d., 1.8 μm; Waters, United States). Mobile phase A consisted of 95% water: 5% acetonitrile, and mobile phase B consisted of 47.5% acetonitrile: 47.5% isopropanol: 5% water. The flow rate was 0.40 mL/minutes, and the column temperature was 40 °C. The MS/MS conditions were as follows: The electrospray ionization source was operated in positive mode or negative mode. The source temperature was 400 °C, the sheath gas flow rate was 40 arb, the auxiliary gas flow rate was 10 arb, and the ion-spray voltage was -2800 V in negative mode and 3500 V in positive mode. The collision energy was normalised, and 20-40-60 V rolling was used for MS/MS. The full MS resolution was 70000, and the MS/MS resolution was 17500. The detection was carried out over a mass range of 70-1050 m/z[23].

Identification of the XQG constituents absorbed in the blood

In addition to metabolomics of the serum taken 2 hours after XQG consumption at follow-up 1 and the fasting baseline in the XQG group, this technology was also applied to the XQG themselves. All ion fragment information obtained from serum and XQG mass spectrometry detection was compared with the MJBIOTCM database, and we included the following data: Constituents present in both XQG samples and the XQG group's follow-up 1 serum, but not detected in their fasting baseline serum; and constituents showing a > 2-fold change in peak intensity in the follow-up 1 serum compared to the fasting baseline serum within the XQG group. If one of the above two points was met, the substance was considered a blood constituent[24].

Network pharmacology and molecular docking verification

TCM systems pharmacology and the encyclopaedia of TCM databases were used to screen the XQG constituents absorbed in the blood and core metabolomic biomarker targets and establish common targets. The maximum correntropy criterion algorithm in the cytoHubba plugin (Cytoscape 3.9.1) was applied to analyze the common targets and discover the key genes. The three dimensional structures of the key genes were identified via the protein data bank database, and the mol2 structures of the XQG constituents absorbed in the blood and the core metabolomic biomarker small molecules in the TCM systems pharmacology database were downloaded. Molecular docking simulations were performed via AutoDock Vina 1.5.7[25]. The docking results were evaluated on the basis of a binding energy less than -5.0 kcal/mol as the criterion for good binding between the active substance and the receptor protein. We selected one XQG constituent absorbed in the blood and one core metabolomic biomarker on the basis of having the lowest total binding energy. The above two substances were combined with the receptor protein that exhibited the strongest binding via PyMOL 3.1.0 for visualization.

Statistical analysis

Statistical analysis was performed using SPSS 23.0 software (International Business Machines Corporation, United States). For continuous variables, the results are presented as the means ± SD. Comparisons were made using the Mann-Whitney U test for nonnormally distributed data or independent/paired-samples t tests for normally distributed data. The trends in the different periods of FBG and 2hPG in the two groups were analyzed using repeated-measures one-way analysis of variance. For the categorical variables, the results are presented as n and percentages, and comparisons between the groups were performed using the χ² test or Fisher’s exact test. Intention-to-treat (ITT) and per-protocol (PP) analyses of the variables were performed because the ITT analyses can prevent biased results due to nonrandom dropout of the participants. The data of all the trial participants can be analyzed at the same time to reflect the real-world treatment effect, which is conducive to the promotion of the scheme in the future. PP analysis includes the participants who fully comply with the research protocol, which may overestimate the efficacy of the XQG and lifestyle interventions[1,26]. Missing data were carried forwards via the baseline observation value method (i.e., missing data were filled with the baseline value) or the last observation carried forwards (i.e., the last observation value at a time point during the study period). P < 0.05 was considered to indicate statistical significance.

Orthogonal partial least squares discriminant analysis (OPLS-DA) was used to display classification changes and differentially abundant metabolite analysis with the ropls R package (v 1.6.2). The Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways were predicted via KEGG keg v20230830. KEGG functional enrichment was generated via SciPy Python (v 1.0.0). An analysis of functional differences was performed via the Wilcoxon test. Receiver operating characteristic (ROC) curves and areas under the curve (AUC) were calculated via the pROC R package (v 1.12.1). Correlation network diagrams were generated via R software (v 4.2.1)[27]. Random forest analysis and least absolute shrinkage and selection operator (LASSO) regression analysis were conducted on the differentially abundant metabolites related to 2hPG. Random forest analysis was performed with sklearn Python (v 0.19.1). LASSO regression analysis was performed with the glmnet R package (v 4.1.7). The discovery of core metabolomic biomarkers was based on common metabolites via random forest analysis and LASSO regression analysis, with coefficients ≥ 10.0.

RESULTS

Study participants

Forty PDM participants were recruited and randomly allocated to the Con group (n = 20) or the XQG group (n = 20). Three participants in the Con group withdrew. In the Con group, one participant did not respond for additional follow-up. Another two participants presented abnormally elevated fasting plasma glucose (FPG) or 2hPG after the first month. One person developed a skin allergy, and one experienced gastrointestinal discomfort after taking XQG for the first time. Owing to mild adverse reactions, both individuals completed the study. The consolidated standards of reporting trials flowchart for this clinical trial is shown in Figure 1.

Figure 1 Consolidated standards of reporting trials participant flow chart. PDM: Prediabetes mellitus; Con: Control; XQG: Xiaokeqing granule; ITT: Intention to treat; PP: Per-protocol.

XQG constituents absorbed in the blood

A total of 2357 constituents were identified in the XQG samples. A total of 1971 constituents were identified in the baseline fasting serum of the XQG group, and 1994 constituents were identified in the 2 hours after XQG consumption at follow-up 1. Fifty-two constituents of XQG were identified as the main constituents absorbed in the blood. The main constituents related to PDM in XQG included artemisinin, oestradiol, groenlandicine, isomangiferin, jatrorrhizine, kaempferol, and palmatine (Figure 2).

Figure 2 Identification of Xiaokeqing granule constituents absorbed in the blood. A: Positive-ion chromatograms of the serum of the Xiaokeqing granule group; B: Negative ion chromatograms of the serum of the Xiaokeqing granule group. TIC: Total ion chromatogram.

Baseline characteristics

The baseline characteristics of the PDM participants in the Con and XQG groups were shown in Table 1 with ITT analysis and PP analysis. There were no statistically significant differences between the two groups (P > 0.05).

Table 1 Comparison of the baseline characteristics, mean ± SD.

Variables	Con group (n = 20, ITT analysis)	Con group (n = 17, PP analysis)	XQG group (n = 20, ITT or PP analysis)
Sex (male/female)	8/12	7/10	5/15
Age (year)	51.10 ± 10.88	52.65 ± 10.52	51.65 ± 12.56
DD (month)	52.25 ± 53.32	56.18 ± 56.39	33.55 ± 33.35
BMI (kg/m2)	23.15 ± 3.09	23.06 ± 3.24	23.64 ± 4.08
SBP (mmHg)	120.85 ± 12.71	120.18 ± 13.50	119.30 ± 15.20
DBP (mmHg)	77.50 ± 8.68	76.65 ± 8.93	79.05 ± 8.74
FPG (mmol/L)	6.48 ± 0.47	6.48 ± 0.49	6.26 ± 0.47
2hPG (mmol/L)	10.20 ± 3.34	10.42 ± 3.51	9.71 ± 2.03
HbA1c (%)	5.97 ± 0.48	5.95 ± 0.31	5.92 ± 0.33
FINS (μIU/mL)	9.23 ± 4.90	9.52 ± 5.27	15.07 ± 12.46
HOMA-IR	2.67 ± 1.47	2.76 ± 1.59	4.15 ± 3.28
ALT (TU/L)	22.15 ± 9.90	23.22 ± 10.22	26.76 ± 17.52
AST (U/L)	21.83 ± 4.43	22.29 ± 4.62	24.99 ± 7.44
BUN (mg/dL)	5.03 ± 1.20	5.19 ± 1.22	5.61 ± 0.96
Cr (mg/dL)	61.50 ± 10.60	61.14 ± 10.85	64.46 ± 15.71

BMI: Body mass index; DD: Disease duration; 2hPG: 2-hour postprandial glucose; HbA1c: Glycated haemoglobin A1c; FINS: Fasting insulin; HOMA-IR: Homeostasis model assessment-insulin resistance; ALT: Alanine aminotransferase; AST: Aspartate aminotransferase; BUN: Blood urea nitrogen; Cr: Creatinine; SBP: Systolic blood pressure; DBP: Diastolic blood pressure; Con: Control; XQG: Xiaokeqing granule; ITT: Intention-to-treat; PP: Per-protocol; FPG: Fasting plasma glucose.

Effectiveness

Compared with the Con group with ITT analysis and PP analysis, the FPG and 2hPG at follow-up 1 were not significantly different in the XQG group (P > 0.05). The 2hPG significantly differed between the Con and XQG groups at Follow-up 2 (P < 0.05). HbA1c only has differences in the PP analysis (P < 0.05). However, FPG, FINS, HOMA-IR, BMI, SBP and DBP were not significantly different (P > 0.05) (Table 2).

Table 2 Comparison of blood glucose and related indicators, mean ± SD.

Variables	Con group (ITT analysis)			Con group (PP analysis)			XQG group (ITT or PP analysis)
	Baseline (n = 20)	Follow-up 1 (n = 20)	Follow-up 2 (n = 20)	Baseline (n = 17)	Follow-up 1 (n = 17)	Follow-up 2 (n = 17)	Baseline (n = 20)	Follow-up 1 (n = 20)	Follow-up 2 (n = 20)
FPG (mmol/L)	6.48 ± 0.47	6.61 ± 0.66	6.47 ± 0.74	6.48 ± 0.49	6.63 ± 0.69	6.46 ± 0.78	6.26 ± 0.47	6.25 ± 0.73	6.09 ± 0.64
2hPG (mmol/L)	10.20 ± 3.34	9.69 ± 2.54	9.77 ± 2.71	10.42 ± 3.51	9.83 ± 2.45	9.93 ± 2.65	9.71 ± 2.03	9.68 ± 2.73	7.98 ± 2.17a,1,2
HbA1c (%)	5.97 ± 0.48	NA	6.14 ± 0.64	5.95 ± 0.31	NA	6.15 ± 0.56	5.92 ± 0.33	NA	5.84 ± 0.29a,2
FINS (μIU/mL)	9.23 ± 4.90	NA	9.97 ± 7.80	9.52 ± 5.27	NA	10.40 ± 8.41	15.07 ± 12.46	NA	14.05 ± 14.49
HOMA-IR	2.67 ± 1.47	NA	2.86 ± 2.27	2.76 ± 1.59	NA	2.98 ± 2.45	4.15 ± 3.28	NA	3.78 ± 3.88
BMI (kg/m2)	23.15 ± 3.09	NA	23.01 ± 3.20	23.06 ± 3.24	NA	22.89 ± 3.35	23.64 ± 4.08	NA	23.16 ± 3.83
SBP (mmHg)	120.85 ± 12.71	NA	119.80 ± 9.65	120.18 ± 13.50	NA	119.47 ± 10.43	119.30 ± 15.20	NA	120.40 ± 12.05
DBP (mmHg)	77.50 ± 8.68	NA	78.95 ± 9.77	76.65 ± 8.93	NA	79.29 ± 10.35	79.05 ± 8.74	NA	77.65 ± 7.20

^aP < 0.05.

¹P vs control group (intention-to-treat analysis).

²P vs control group (per-protocol analysis).

Con: Control; XQG: Xiaokeqing granule; ITT: Intention-to-treat; PP: Per-protocol; FPG: Fasting plasma glucose; 2hPG: 2-hour postprandial glucose; HbA1c: Glycated haemoglobin A1c; FINS: Fasting insulin; HOMA-IR: Homeostasis model assessment-insulin resistance; BMI: Body mass index; SBP: Systolic blood pressure; DBP: Diastolic blood pressure; NA: Not analyzed.

Safety

There were no adverse reactions in the Con group, and there were 2 participants in the XQG group. There was no significant difference between the two groups (P > 0.05) (Table 3). Moreover, there were no significant differences in liver or renal function after the intervention (P > 0.05) (Table 4).

Table 3 Adverse events comparison, n (%).

Group	Number of participants number (ITT or PP analysis)	Number of AE
Con group	20/17	0 (0)
XQG group	20/20	2 (10.0)

Con: Control; XQG: Xiaokeqing granule; ITT: Intention-to-treat; PP: Per-protocol; AE: Adverse events.

Table 4 Comparison of liver and renal function, mean ± SD.

Variables	Con group (ITT analysis)		Con group (PP analysis)		XQG group (ITT or PP analysis)
	Baseline (n = 20)	Follow-up 2 (n = 20)	Baseline (n = 17)	Follow-up 2 (n = 17)	Baseline (n = 20)	Follow-up 2 (n = 20)
ALT (U/L)	22.15 ± 9.90	23.61 ± 18.73	23.22 ± 10.22	24.94 ± 20.01	26.76 ± 17.52	26.65 ± 18.95
AST (U/L)	21.83 ± 4.43	23.01 ± 8.30	22.29 ± 4.62	23.68 ± 8.85	24.99 ± 7.44	26.94 ± 14.01
BUN (mmol/L)	5.03 ± 1.20	5.18 ± 1.49	5.19 ± 1.22	5.37 ± 1.54	5.61 ± 0.96	5.66 ± 1.32
Cr (μmol/L)	61.50 ± 10.60	59.70 ± 9.81	61.14 ± 10.85	59.02 ± 9.80	64.46 ± 15.71	63.84 ± 15.24

Con: Control; XQG: Xiaokeqing granule; ITT: Intention-to-treat; PP: Per-protocol; ALT: Alanine aminotransferase; AST: Aspartate aminotransferase; BUN: Blood urea nitrogen; Cr: Creatinine.

Dropout rate

In this study, a total of 3 participants were dropped, accounting for 7.5% of the sample. All 3 participants were from the Con group. There were no significant differences between the Con and XQG groups (P > 0.05) (Table 5).

Table 5 Comparison of the dropout rates, n (%).

Group	Participants (n)	Dropped out
Con group	20	3 (15.0)
XQG group	20	0 (0)

Con: Control; XQG: Xiaokeqing granule.

Metabolites

The ion peak shape and separation of the quality sample are good, indicating that the analysis system is stable. The z score of the internal standard is between -2 and 2. This indicates that the sample processing is stable. These factors can ensure the reliability of the analysis results. In total, we identified 940 positive-ion and 1145 negative-ion serum metabolites. Among them, 328 positive ions and 396 negative ions were found among the KEGG metabolites. The serum positive-ion and negative-ion metabolic data of the Con and XQG groups were separated by an OPLS-DA model (Figure 3A and B). The differentially abundant metabolites identified in the volcano plots included 129 upregulated genes and 11 downregulated genes (Figure 3C). The most significant difference among them was that of rhein (P = 0.0001041). In terms of the differentially abundant metabolites, the significantly enriched KEGG topology pathways included tryptophan metabolism, pantothenate and certificate of analysis (CoA) biosynthesis, lysine degradation and cofactor biosynthesis (Figure 3D). The relationships between metabolites and KEGG signalling pathways are mutual. For example, 5-hydroxy-L-tryptophan, L-formylkynurenine and 3-methylindole are related to tryptophan metabolism, and 3-methyl-2-oxobutanoic acid and 4’-phosphopantothenoylcysteine are related to pantothenate and CoA biosynthesis (Figure 3E and Table 6).

Figure 3 Comparisons of metabolites between the control group and Xiaokeqing granule group. A: Comparison of the positive-ion metabolite profiles by orthogonal partial least squares discriminant analysis (OPLS-DA); B: Comparison of the negative ion metabolite profiles by OPLS-DA; C: Volcano plot of the significantly differentially abundant metabolites; D: Metabolic Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway topology analysis; E: Metabolic KEGG pathway enrichment analysis network. Green circles represent differentially metabolites, orange and yellow triangles represent differentially KEGG pathways, and the size represents the number of metabolites in the pathways. Con: Control; XQG: Xiaokeqing granule; FC: Fold change; VIP: Variable importance in projection; CoA: Certificate of analysis.

Table 6 Metabolite and Kyoto Encyclopedia of Genes and Genomes pathways.

Metabolite	KEGG pathway description
5-Hydroxy-L-Tryptophan	Tryptophan metabolism
L-Formyl kynurenine	Tryptophan metabolism, Biosynthesis of cofactors
2,3,4,5-tetrahydro-2-pyridinecarboxylic acid	Lysine degradation
N6-Acetyl-L-Lysine	Lysine degradation
Phenethylamine glucuronide	Biosynthesis of cofactors
3-Methyl-2-oxobutanoic acid	Pantothenate and CoA biosynthesis, Biosynthesis of cofactors
3-Methylindole	Tryptophan metabolism
4’-Phosphopantothenoylcysteine	Pantothenate and CoA biosynthesis, Biosynthesis of cofactors

KEGG: Kyoto Encyclopedia of Genes and Genomes; CoA: Certificate of analysis.

Potential biomarkers

Among the 140 differentially abundant metabolites at baseline, the only 5 biomarkers identified via random forest analysis that were related to 2hPG included 5-hydroxy-L-tryptophan, carnosol, lanthionine ketimine, N-nervonoyl methionine and N-oleoyl glutamine (Figure 4A and Table 7). The 15 important biomarkers included 5-hydroxy-L-tryptophan, N-nervonoyl methionine, calcitroic acid, N-oleoyl glutamine, valyl asparagine, lanthionine ketimine and 4-epitetracycline, among others, according to LASSO regression analysis (Figure 4B, Tables 7 and 8). Common biomarkers included N-nervonoyl methionine, lanthionine ketimine, 5-hydroxy-L-tryptophan and N-oleoyl glutamine. The core metabolomic biomarkers were N-nervonoyl methionine and 5-hydroxy-L-tryptophan. The AUCs of the core biomarkers were 0.765 and 0.780, respectively (Figure 4C). When we combined the 2 biomarkers for diagnosis, the AUC increased to 0.843, and the diagnostic efficacy of the model was good (Figure 4D).

Figure 4 Comparisons of potential biomarkers between the control group and Xiaokeqing granule group. A: Important biomarkers according to random forest analysis; B: Important biomarkers according to least absolute shrinkage and selection operator regression analysis; C: Receiver operating characteristic analysis of the two core metabolomics biomarkers; D: Receiver operating characteristic curve analysis of the joint core metabolomics biomarkers. TPR: True positive rate; FPR: False positive rate; AUC: Area under the curve; CI: Confidence interval.

Table 7 Evaluation index table of the total feature model.

Method	Kappa	F1	Precision	Recall	Accuracy
Random forest	0.7843	0.8889	0.8421	0.9412	0.8919
LASSO	0.6667	0.8333	0.8333	0.8333	0.8333

LASSO: Least absolute shrinkage and selection operator.

Table 8 Metabolite features of least absolute shrinkage and selection operator regression analysis.

Feature	Coefficient
Dimethylformamide	-0.36107
5-Hydroxy-L-Tryptophan	10.6295
Lanthionine ketimine	0.750093
5-Nitro-2-phenylpropylaminobenzoic acid	0.214502
Tyr Leu	0.242158
N-Nervonoyl methionine	12.02622
Calcitroic acid	4.998331
Valylasparagine	1.677792
4-Epitetracycline	0.702722
Naringenin 7-sulfate	0.512388
Stearoyllactic acid	-1.58267
N-Oleoyl glutamine	2.505959
1-Nitro-7-glutathionyl-8-hydroxy-7,8-dihydronaphthalene	0.374231
2-(5-carboxypentanoylamino) benzoic acid	0.332139
Flurenol	0.023748

Network pharmacology and molecular docking

The present study identified 356 targets from the 7 XQG constituents absorbed in the blood and 49 targets from the 2 core metabolomic biomarkers. Among them, 21 common targets are shown in Table 9. The key genes included HTR1A, SLC6A4, HTR2C, HTR2B, SLC6A3, and HTR5A (Figure 5A). The molecular docking results revealed that key receptor proteins have relatively stable binding abilities with the XQG constituents absorbed in the blood and core metabolomic biomarkers (Figure 5B). The total binding energies of the 9 substances are -48.7, -34.3, -52.3, -48.7, -42.9, -51, -47.5, -41.9, and -42.3 kcal/mol (Figure 5C). Groenlandicine maintains a stable conformation with HTR2C through hydrogen bonding with LEU-209. 5-Hydroxy-L-tryptophan maintains a stable conformation with SLC6A4 through hydrogen bonding with aspartic acid-121 and SER-204 (Figure 5D).

Figure 5 Network pharmacology and molecular docking. A: Key targets; B: Molecules docking heatmap of Xiaokeqing granule constituents absorbed in the blood, core metabolomics biomarkers and key targets; C: Molecular docking visualization of groenlandicine and HTR2C; D: Molecular docking visualization of 5-hydroxy-L-tryptophan and SLC6A4.

Table 9 Common targets of Xiaokeqing granule constituents and core metabolomic biomarkers.

Targets	Targets	Targets
HTR2B	EGFR	MTNR1A
HTR1A	GRIA1	MTNR1B
HTR2C	ADORA3	MMP3
IDO1	SLC6A3	ITGB1
SLC6A4	MIF	PTPN1
DRD3	FAAH	SIRT2
HTR5A	TRPV1	BACE1

DISCUSSION

This is the first study that compared the effects of XQG and lifestyle interventions in participants with PDM. When participants were included in the XQG group, TCM revealed that the participants had good adherence to the XQG and lifestyle interventions, especially in terms of 2hPG. In addition, we identified potential targets and pathways of XQG and lifestyle action through nontargeted metabolomics techniques.

Effects of XQG and lifestyle on glycaemic control

Elevated levels of FBG, 2hPG and HbA1c are characteristic of patients with PDM. Moreover, these factors are used as the main curative indicators for glycaemic control[28]. Lifestyle changes, such as a healthy diet and increased exercise, have been shown to prevent PDM[29]. This study revealed that XQG combined with lifestyle intervention improved the 2hPG level in the XQG group compared with that in the Con group and that this treatment plan was safe. The application of XQG is in line with the TCM pathogenesis of PDM. XQG promote the balance of the body’s qi, xue, yin, and yang through nourishing the yin and clearing heat to regulate blood glucose. This finding is consistent with previous TCM research[29], including studies on Jinlida granules, Danzhi Jiangtang capsules, Tian-Huang Formulae, Tianqi capsules, Shenzhu Tiaopi granules, etc.[14,19,30-32]. Moreover, the package insert of XQG mentions that the course of treatment is 2 months, so we chose 2 months as an observation cycle. Although the treatment period was relatively short, this study showed that the 2hPG level decreased in the participants, instilling confidence that blood glucose levels in PDM patients can be controlled. In this study, the effect on HbA1c reduction was significant in the PP analysis. However, this result was not replicated in the ITT analysis. Because HbA1c reflects three-month blood glucose levels, this phenomenon might be explained by the short observation period in this study. The effect on FBG reduction was nonsignificant. This might be because although the FBG was increased, the degree of increase was very small. Therefore, after the XQG and lifestyle interventions, the decrease in blood glucose was also small and thus nonsignificant. In addition, we observed that the HOMA-IR in the XQG group showed a downward trend, although there was no significant difference. We will focus these aspects in the later stage.

The improvement in 2hPG may be related to the absorption of some XQG constituents in the blood. Artemisinin can significantly reduce FBG, 2hPG and HbA1c by promoting insulin secretion, protecting pancreatic islet beta cells, attenuating insulin resistance and regulating the phosphatidylinositol 3-kinase (PI3K)/protein kinase B (AKT) pathway[33-35]. The incidence rate of T2DM in women after menopause increases, which is related to the decrease in oestrogen in their bodies. Oestradiol promotes the regulation of insulin secretion, insulin biosynthesis and sensitivity in pancreatic islet beta cells[36]. In addition, oestradiol can restore the balance of the gut microbiome to decrease blood glucose[37]. The phytoestrogen resveratrol can improve insulin, protect pancreatic islet beta cells, and increase the level of insulin receptor substrate 1 (IRS1)[38]. Groenlandicine and jatrorrhizine are derived mainly from Coptis chinensis Franch and are attributed to quaternary protoberberine-type alkaloids. Groenlandicine exhibited moderate antidiabetic effects[39]. The reason for this effect may be that it can specifically bind to α-glucosidase[40]. Groenlandicine has good binding activity with HTR2C. A lack of HTR2C is associated with increased food intake, leading to impaired glucose tolerance and weight gain[41,42]. Groenlandicine may have reduced hyperphagia, thereby preventing weight gain and improving glucose tolerance[43]. These findings indicate a research direction for the analysis of the effects and underlying mechanism of XQG and lifestyle interventions. Jatrorrhizine exerts antidiabetic effects by inhibiting human islet amyloid polypeptide to improve glucose sensitivity[44,45]. By inhibiting aldose reductase or enhancing glucose uptake, isomangiferin, similar to polyphenols and glycosidic flavonoids, can improve glucose tolerance in db/db diabetic model mice[46-48]. Isomangiferin is derived mainly from Anemarrhena asphodeloides Bunge in XQG. Kaempferol is derived mainly from Typha angustifolia L., and Euphorbia humifusa Willd is a type of flavonoid. It has antidiabetic effects by regulating the nuclear factor kappa-B, adenosine 5’-monophosphate-activated protein kinase, and PI3K/AKT pathways or by increasing binding affinities for dipeptidyl peptidase-4 and sodium-glucose cotransporter-1[49-51]. Palmatine is derived from Coptis chinensis Franch and can decrease FBG and increase insulin in T2DM mice. This may regulate the IRS1/RAC-beta serine/AKT2/FOXO1/glucose transporter 2 pathway or inhibit lipase and cholinesterases[52-54].

Effects of XQG and lifestyle on serum metabolites

These results revealed significant overall differences between the Con and XQG groups through OPLSDA analysis. A total of 140 differentially abundant metabolites were screened, among which 10 main differentially abundant metabolites are listed in the volcano map. Rhein can reduce FBG levels and improve glucose tolerance in individuals with T2DM by ameliorating insulin resistance and protecting pancreatic islet cells[55,56]. Oleoyl glutamine can cause mice to exhibit the HOMA-IR phenotype[57]. N-Oleoyl glutamine is a potential biomarker of the effects of XQG and lifestyle on PDM. The AUC of N-nervonoyl methionine was 0.765, indicating that it is also a biomarker. However, there is very little discussion about its role in PDM. It has similar nutritional and metabolic effects as L-methionine. L-methionine has been administered as a novel therapeutic method for preventing islet beta-cell death caused by hyperglycaemia[58]. N-acetyl-L-tyrosine, a PPARA activator, can improve insulin sensitivity. N-Acetyl-L-tyrosine has potential treatment effects on T2DM[59]. The NLRP3 inflammasome plays an important role in the development of T2DM. Lamivudine can improve insulin sensitivity and reduce inflammasome activation to prevent T2DM[60]. In T2DM model rats, the plasma levels of Gly Tyr decreased. XQG and lifestyle can increase the level of Gly Tyr to improve T2DM[61]. The hypoglycaemic effect of tryptophan is mediated through intracellular 5-hydroxytryptamine. 5-Hydroxy-L-tryptophan was shown to decrease in the pancreas of T2DM rats[62]. These results suggest that 5-hydroxy-L-tryptophan induces more rapid hypoglycaemia[63]. One of the targets of XQG and lifestyle interventions is 5-hydroxy-L-tryptophan, as indicated by an AUC of 0.780. On the basis of machine learning algorithms, we jointly analysed N-nervonoyl methionine and 5-hydroxy-L-tryptophan and derived an AUC of 0.843, indicating that this combination is an important target for the effectiveness of XQG and lifestyle interventions. The relationships among N-nervonoyl methionine, 5-hydroxy-L-tryptophan and PDM with yin deficiency syndrome with excessive heat will be explored next.

In our study, we found that tryptophan metabolism plays an important role in PDM. Reports in the literature on tryptophan metabolism have revealed a strong relationship with the progression of T2DM[64]. The tryptophan metabolic pathway is involved in islet beta cell function, HOMA-IR, the gut microbiota, and the intestinal barrier[65]. Metformin corrected the metabolic disorders associated with tryptophan metabolism. These findings provide new evidence for the mechanism by which metformin regulates blood glucose[66]. Dapagliflozin is an antidiabetic drug that functions via the tryptophan metabolism-glucagon-like peptide 1 axis and participates in islet beta cell regeneration to lower blood glucose[67]. The relationships between metabolites and KEGG signalling pathways are mutual. 5-Hydroxy-L-tryptophan is involved in tryptophan metabolism. In our research, 5-hydroxy-L-tryptophan was shown to have good binding activity with SLC6A4. SLC6A4 expression levels are significantly associated with hyperglycaemia risk[68,69]. Among the study genes, SLC6A4 was one of the most differentially expressed genes, making it one of the possible targets of XQG and lifestyle for regulating hyperglycaemia. HTR1A, which encodes the 5-Hydroxytryptamine receptor 1A, has an important influence on women with diabetes[70]. Leech and centipede granules can improve T2DM complications by affecting this gene[71]. In addition, HTR1A expression levels are associated with type 1 diabetes mellitus risk[72].

Pantothenate and CoA biosynthesis are correlated with glucose metabolism[73]. A series of studies have shown that heat-treated foxtail millet protein, camel milk peptides, mangiferin calcium salt, and C. moschata can decrease FBG and alleviate HOMA-IR in PDM or T2DM mice[74-77]. These studies suggest that these changes may be related to the regulation of pantothenate and CoA biosynthesis. Chronic low-grade inflammation is a type of T2DM syndrome. Lysine degradation plays an important role in primary immunometabolism[78]. Berberine improves blood glucose, HbA1c, and glucose tolerance by multiple targets or multiple pathways. One of these pathways involves lysine degradation[79]. Cofactor biosynthesis is an important foundation for the function of mitochondria[80]. Mitochondrial dysfunction is involved in the development of insulin resistance and beta-cell failure[81]. For example, one study showed that various thiamines used as cofactors were present at low levels in patients with T2DM[82].

There are several limitations in this study. The sample size was small; therefore, this study was a preliminary investigation and an exploratory randomised controlled trial. There were only significant differences in 2hPG. The prolonged effect of XQG and lifestyle interventions on the prognosis of PDM was not observed because of the short follow-up duration. On the basis of the effectiveness of this research, we will carry out longitudinal studies with large sample sizes in the future to provide higher-level evidence for the prevention and treatment of PDM via TCM. This study is a single-center clinical trial, and multiple different regions still need to be included so that the study protocol can be directly extended to populations from other regions. The lack of an XQG placebo in this study is a limitation, and the placebo effect may have been overlooked, which may have caused some bias in the results. In future studies, we will use a placebo as a control to ensure that the results of the study are evaluated more objectively. The selection of the XQG constituents absorbed in the blood may cause some deviation, as we used samples taken one month after taking XQG instead of 7 days. In the future, we will choose the classic method, i.e., a 7-day follow-up, for greater standardization of the analysis of the XQG constituents absorbed in the blood. Furthermore, we selected the target of action on the basis of a literature review and network pharmacology analysis. Future research could include the application of proteomics or transcriptomics methods to determine candidate XQG and lifestyle interventions targets.

CONCLUSION

XQG and lifestyle interventions improved 2hPG in participants with PDM. The main constituents absorbed in the blood were artemisinin, oestradiol, and groenlandicine jatrorrhizine. The core metabolomic biomarkers are N-nervonoyl methionine and 5-hydroxy-L-tryptophan. The mechanism of this effect is related to improved tryptophan metabolism, pantothenate and CoA biosynthesis, etc. Further research is needed to provide evidence that XQG and lifestyle interventions promote long-term glycaemic control and to explore the mechanisms involved.