Objective studies were identified through searching for papers published between January 2007 and March 2024 in the following databases: Web of Science (www.webofscience.com), PubMed (https://pubmed.ncbi.nlm.nih.gov/), Embase (www.embase.com) and China National Knowledge Infrastructure (oversea.cnki.net/index/), without implementing any language restrictions. Broader terms of mesh and free-text terms were applied to identify target drugs and diseases: ‘Astragalus polysaccharide’, ‘Diabetic nephropathies’, ‘Diabetic Glomerulosclerosis’ and ‘Diabetic Kidney Disease’. If necessary, the authors of the identified studies were contacted to obtain more information.

According to the patient/population, intervention, comparison and outcomes principle, the following inclusion criteria were applied: i) Regarding the participants, DN model animals were selected; ii) concerning the intervention, the animals were treated with APS; iii) comparisons were made between the treated animals and the control groups, including untreated controls; and iv) regarding the outcomes, the primary outcome indicators included creatinine (CR), kidney to body weight ratio (KI), blood urea nitrogen (BUN), urine protein (Upro) and fasting blood glucose (FBG).

The exclusion criteria comprised the following: i) Duplicate data or studies; ii) if the article was merely a case report, clinical trial, opinion, abstract, review or an in vitro trial; iii) no control group or treatment group was provided, or the study was performed either without APS, or APS was combined with other drugs; iv) no sample size or specific statistical analysis method was shown; and v) the study did meet the aforementioned inclusion criteria, but the full text was not available (Fig. 2).

All selected documents were imported into Endnote X9, and duplicates were removed. According to the inclusion and exclusion criteria, two researchers performed the literature searches independently. Both the title and the abstract were checked to remove irrelevant literature, and the remaining studies were then carefully examined by reading the full text. The following relevant information from the selected studies was extracted: i) The first author name and the publication date; ii) the basic information, including species, group, body weight (BW) and sample size; iii) the modeling methods and success criteria; iv) the APS intervention route, APS dose and treatment duration; and v) the outcome measures and statistical differences between the different groups. An Excel database was created, and the data were extracted from the target articles manually. The authors of the relevant articles were contacted for their original data when the results were presented graphically only; if no response was received from the authors concerned, then WebPlotDigitizer 4.5 software (https://automeris.io/WebPlotDigitizer) was used to quantify the graphic data. The last data point among the multiple time points was subsequently extracted for meta-analysis. If multiple doses of APS were applied in the treatment groups, the highest group was selected for analysis. The formula ‘SD=SEM x n1/2’ was employed to convert the data into SD where the data in the text were shown as SEM (26). The SYRCLE risk of bias tool for animal studies was applied to examine the risk of bias in the included studies by two assessors independently (27). The assessments of measurement bias, selection bias, reporting bias, implementation bias and other sources of bias were included. Any disagreements were resolved with the corresponding author during the quality evaluation.

The statistical software package STATA (15.1 edition; StataCorp LP) was applied for the statistical analyses. It was ensured that the results included were continuous variables; therefore, the total effect size was expressed in terms of standardized mean differences (SMDs) and 95% confidence intervals (CIs). P<0.05 of was considered to indicate a statistically significant difference. The statistical heterogeneity was assessed using I-Square (I²) analysis. If I²>50% was considered to indicate statistically significant heterogeneity, the random effects model was used, whereas the fixed effects model was used in all other cases (28). To explore the underlying causes of heterogeneity, subgroup and sensitivity analyses were performed, including the indicators of FBG, Cr, BUN and Upro. To perform the subgroup analysis, subgroups were designated according to the following: i) the species (mice or rats); ii) the treatment duration (≤8 weeks or >8 weeks); and iii) the APS dose (<400 mg/kg/day or ≥400 mg/kg/day). Egger's linear regression analysis and Begg's rank correlation analysis were applied to evaluate the publication bias of FBG, Cr, BUN and Upro. If publication biases were to be identified, then trim-and-fill methods were applied. In order to better present the effects of intervention duration and the APS dose on the results, time-dose response relationship plots of FBG, Cr, BUN and Upro were created. All groups with results P<0.05 were included in the time-dose analysis for the multiple groups that participated in the same study.

Conducting a search of the four major databases (Web of Science, PubMed, Embase and CNKI), the existence of 179 potential target articles was disclosed, including 59 from PubMed, 14 from Web of Science, 22 from Embase and 84 from CNKI. After consolidating the 179 articles and removing the duplicates, 93 records were retained. After performing two more screenings according to the pre-set criteria, a final total of 8 articles met the eligibility requirements, and were finally selected in the current meta-analysis. The selection process of the articles is shown in Fig. 2.

A total of 8 studies were included, and the DN models were based on either male rats or mice. In total, 226 animals were used, including 82 in the model group and 144 in the treatment group. A total of 6 studies used Sprague Dawley rats (184/226; 81.4%), 1 study used Wistar rats (30/226; 13.3%), and 1 study used db/db mice (12/226,5.3%). The animal weight was mentioned in five of the studies, and age was mentioned in six of the studies. Different doses of streptozotocin (STZ) were applied during the construction of the DN model in 7 studies, among which three of those studies used STZ combined with a high sugar, high fat diet, with high fat feed to replicate the conditions of diabetes. Spontaneous diabetic model (SDM) mice were used in 1 study. All studies considered FBG >11.1 mmol/l (range: 11.1~16.7 mmol/l) as the criterion to evaluate the success of the models, with the exception of the study that used SDM model mice. The minimum and the maximum duration of administration of APS were 6 and 12 weeks, respectively, whereas the minimum and the maximum doses of APS were 25 mg/kg/day and 1,000 mg/kg/day, respectively. Regarding the primary outcome measures, 6 studies recorded FBG levels, 7 studies recorded BUN levels, 5 studies recorded Cr levels, and 6 studies recorded Upro levels. A total of 3 studies focused on inflammatory indicators, including IL-6, IL-1β and TNF-α. Certain studies mentioned markers of oxidative stress, such as malondialdehyde (MDA) and superoxide dismutase (SOD). The detailed information concerning APS is shown in Table I, and the features of the present study are shown in Table II.

As shown in Fig. 3, the quality of the articles was evaluated strictly according to the aforementioned criteria. The Jadad score results ranged from 5-7 points. One study (29) received 5 points, 2 studies (18,20) received 6 points, and 5 studies (19,21,30-32) received 7 points each (Table III). In all the included studies, animals were grouped using a randomized method. All studies demonstrated baseline characteristics between the APS groups and the DN group. The experimental setting was identical, and none of the studies described whether the distribution of different groups was adequately masked. In addition, whether or not the grouping and setting of animals conformed to the principle of randomization was also considered in the present study. Furthermore, all 8 articles were randomized and described results with complete data, and no selective reporting bias was observed. No other sources of bias were identified in the present meta-analysis (Fig. 3).

All 8 articles presented the FBG data. In all cases, the APS intervention group led to a significant improvement in the FBG level [n=156; SMD: -4.029 (95% CI: -4.929 to -3.13), P<0.05; heterogeneity: I²=58.5%, P<0.05; Fig. 4].

Effect of APS intervention on Cr. All 8 articles presented the Cr data, and the results of all these studies revealed that the APS intervention group led to a significant improvement in the Cr level [n=156; SMD: -5.037 (95% CI: -6.353 to -4.26), P<0.05; heterogeneity: I²=53%, P<0.01; Fig. 5].

Effect of APS intervention on BUN. All 8 articles presented the BUN data, and the results of all these studies showed that the APS intervention group led to a significant improvement in the BUN level [n=156, SMD: -3.13 (95% CI: -3.954 to -2.306), P<0.01; heterogeneity: I²=63.4%, P<0.01; Fig. 6].

Effect of APS intervention on Upro. All 8 articles presented the Upro data, and the results of all these studies demonstrated that the APS intervention group led to a significant improvement in the Upro level [n=156; SMD: -3.241 (95% CI: -4.479 to -2.003), P<0.01; heterogeneity: I²=84.2%, P<0.01; Fig. 7].

Effect of APS intervention on KI. Usually, the kidney-to-body weight ratio remains relatively stable (33). An increase in the level of KI is indicative of edema, congestion or hypertrophy in the kidney tissue, whereas a decrease in KI suggests the presence of degenerative changes, such as renal atrophy (34). Data on KI were provided in all studies. The KI increased in the DN model in the 6 relevant articles, and the results identified that APS treatment led to an improvement in the KI level [n=120; SMD: 3.956 (95% CI: 5.825 to 2.087), P<0.01; heterogeneity: I²=88.3%, P<0.01; Fig. 8A]. The KI was found to decrease in the remaining 2 articles that described a DN model, and the results showed that APS treatment led to an improvement in the KI level [n=36; SMD: 1.058 (95% CI: 0.445 1.870), P<0.05; heterogeneity: I²=0%, P>0.01; Fig. 8B]. Considered overall, APS treatment led to an improvement in the KI level of the DN groups.

The indicator of BW was provided in 6 of the articles, which all described a decrease in the BW in the DN model, and the data indicated that APS intervention led to an improvement in the BW level compared with the model group [n=118; SMD: 2.361 (95% CI: 1.545-3.178), P<0.05; heterogeneity: I²=61.4%, P<0.01; Fig. 9].

Effect of APS intervention on total cholesterol (TC). Only 3 articles employed TC as an outcome index. The studies showed a TC increase in the DN model groups, whereas APS intervention groups could improve the TC level (P<0.05). [n=70; SMD: -4.832 (95% CI: -7.394 to -2.27), P<0.01; heterogeneity: I²=83.8%, P<0.01; Fig. 10].

The sensitivity of FBG, BUN, Cr and Upro was subsequently analyzed. After sequentially excluding each study from the meta-analysis, no significant difference was observed between the pre-sensitivity and post-pooled effects of FBG, BUN, Cr and Upro (Fig. 11). Due to the high heterogeneity of each study, FBG, Cr, BUN, Upro and other indicators associated with animal species, the type of DN modeling method, treatment time and APS dose were evaluated. These analyses revealed that the DN modeling method, animal species and APS dose may have provided the source of heterogeneity in FBG and BUN, whereas the DN modeling method and animal species may have provided the source of heterogeneity in Upro and Cr.

The results obtained showed that the publication bias of FBG, Cr, BUN and Upro was statistically significant (P<0.05). The findings also revealed that data from the missing studies did not change the size of the overall combined effect size. The publication bias details are shown in Fig. 12.

The analysis of the included studies demonstrated the potential primary mechanisms of APS treatment on DN (Fig. 13).

Anti-fibrotic and autophagy promotion. Renal interstitial fibrosis is the eventual result of all progressive chronic kidney diseases that lead to end-stage renal disease (35). APS treatment is known to promote autophagy via reducing the protein levels of p62 and mammalian target of rapamycin, while promoting the microtubule-associated protein light chain 3 I/II ratio. Furthermore, APS has been shown to reduce fibrosis by downregulating the expression levels of fibronectin, collagen IV and transforming growth factor-β (TGF-β) (18). The activation of the advanced glycation end products pathway in advanced diabetes mellitus has been identified to be closely associated with the occurrence of DN (36), and activation of AGE receptor can induce activation of the PI3K-Akt, TGF-β/Smad and NF-κB signaling pathways, among others (37). APS was also shown to reduce the expression of the inflammatory factors IL-1β, IL-6 and monocyte chemoattractant protein 1 in DN rats, to inhibit the activity of the Toll-like receptor 4 (TLR4)/NF-κB pathway, and to significantly alleviate kidney injury (21). The PI3K-Akt signaling pathway is involved in multiple cellular processes, including cell proliferation, migration, adhesion and survival (38). APS treatment also has been revealed to decrease the levels of serum inflammatory factors, and to increase the extent of apoptosis of glomerular cells in DN rats, which is associated with inhibition of the PI3K/Akt signaling pathway (32). Ski-related protein N functions as a primary nuclear transcriptional suppressor in the TGF-β/Smad pathway (39). APS treatment has also been shown to influence the TGF-β/Smad signaling pathway, leading to the protection of renal function (19). Activation of aldose reductase can trigger the activation of protein kinase C, mitogen-activated protein kinase, and other signaling pathways, resulting in the overexpression of TGF-β, TNF-α and other cytokines, leading to a variety of pathophysiological chain reactions associated with DN (40).

The relief of inflammation. The chemokines and cytokines that participate in inflammatory processes have been shown to be closely associated with the occurrence and development of DN (41). NF-κB is a downstream effector of the TLR4 signaling pathway that mediates various inflammatory processes (42). APS has been demonstrated to ameliorate DN renal injury through inhibiting the TLR4/NF-κB signaling pathway, and thereby alleviating the inflammatory response. In addition, APS treatment was also found to increase the expression of IκB mRNA and decrease the level of NF-κB mRNA in the renal cortex (20).

The relief of oxidative stress. Oxidative stress is one of the most important factors in the occurrence and development of DN (43). AMP-activated protein kinase (AMPK), as a metabolic regulator and energy sensor, has an important role in the entire pathophysiological process of DN (44). It has been reported that AMPK can affect the expression and phosphorylation process of peroxisome proliferator-activated receptor gamma coactivator 1α, thereby influencing the production of cell energy (45). APS treatment has been revealed to downregulate the AMPK/SIRT1/FOXO1 signaling pathway in DN rats, and to activate autophagy, inhibit oxidative stress and alleviate kidney injury (30).

Limitations and considerations. In general, demonstrating the effects of therapeutic drugs through performing animal experiments is an important requirement in the preclinical research and development of new drugs. However, problems with preclinical animal studies may arise from a high risk of experimental bias and the low reproducibility of results. These lead to low success rates for the development of safe and effective drugs in clinical trials (46). Along these lines, the systematic evaluation of the present study has the following shortcomings. First, only 8 high-quality studies were identified and screened, which may inevitably lead to a bias. Secondly, the data extracted from certain of these articles were obtained indirectly through data extraction software, which may have led to measurement bias. Thirdly, the high heterogeneity of the FBG, BUN, Cr, Upro and KI values cannot be ignored in spite of the subgroup analysis. This heterogeneity may have caused the calculated results to differ from the actual values. The current experimental conclusions are mainly based on male animals, and the lack of female mice is a limitation, thus their applicability to females remains uncertain until further studies with female subjects are conducted. While suggesting APS could improve DN potentially via relieving inflammatory responses and attenuating the TLR4/NF-κB signaling pathway, it is acknowledged that basing the current findings on only male mice is a limitation as sex differences may influence these mechanisms, warranting future studies with female mice for a more comprehensive understanding.

In conclusion, the data obtained to date have suggested that treatment with APS leads to improvements in renal function and proteinuria in DN animal models, and that this has a role in reducing the burden of renal dysfunction secondary to diabetes. According to our data, the protective mechanisms of APS on DN may be associated with improving the inflammatory status, fibrosis degree and oxidative stress status of DN model animals. In order to more accurately evaluate the DN efficacy and safety of APS, however, larger and more long-term studies of higher quality are required to confirm these findings prior to clinical application.