To identify pertinent animal studies, two independent authors searched 10 databases, including PubMed, Embase, Web of Science, Scopus, SCIELO, Cochrane Library, CNKI, Wanfang Database, CBM and VIP, covering records from database inception through November 2024, with language restrictions applied to English and Chinese publications. Detailed search strategies are provided in Supplementary Table S1.

Specified inclusion criteria established beforehand were: (1) Participants: animal models of renal fibrosis exhibiting pathological histological changes consistent with renal fibrosis; (2) Intervention: The treatment group received varying doses of BBR; (3) Control: The model group received either a placebo or no treatment; (4) Outcomes: The principal outcome indicators included serum creatinine (Scr), blood urea nitrogen (BUN), transforming growth factor-β1 (TGF-β1), and α-smooth muscle actin (α-SMA). Further outcome indicators included 24-h urinary protein, kidney weight index (KWI), renal fibrosis area, kidney injury molecule-1 (KIM-1), renal injury score, Collagen IV, Collagen I, fibronectin (FN), E-cadherin, superoxide dismutase (SOD), malondialdehyde (MDA), NOD-like receptor protein 3 (NLRP3), glutathione peroxidase (GSH-Px), and interleukin-1β (IL-1β).

The pre-specified exclusion criteria were: (1) studies focusing exclusively on cellular models or clinical research; (2) non-renal fibrosis models; (3) redundant or overlapping publications; (4) reviews; (5) conferences proceedings, dissertations, and thesis presentations; (6) case reports; (7) incomplete data; and (8) studies with unavailable full text.

Based on the predefined selection criteria, two independent assessors evaluated the study titles, summaries, and full manuscripts to identify those suitable for inclusion. When discrepancies occurred, we resolved them by discussing with another researcher. A standardized pretest form (Excel) was utilized to obtain relevant information from the selected studies for evidentiary integration. The extracted information included: first author, publication year, fundamental features of the experimental animals (including species, gender, sample size, and weight), intervention specifics (administration route, dose, duration), modeling methodology, histological staining method for renal pathology, outcome measures, and intergroup differences. Data were included exclusively corresponding to the maximum dose of BBR administered. For outcomes with multiple reported time points, we chose to analyze data from the longest observation period. When the data essential for us were presented merely in graphical format, we used Web Plot Digitizer (https://apps.automeris.io/wpd4/) to derive relevant data from the charts.

Two investigators separately evaluated the methodology’s quality in the incorporated studies by means of the SYRCLE tool for animal experiments (Hooijmans et al., 2014). This tool encompasses various types of biases, categorized as follows: 1) Selection Bias, which includes aspects such as the generation of sequences, baseline characteristics, and concealment of allocation; 2) Performance Bias, involving the random assignment of animals and ensuring that caregivers and investigators are blinded; 3) Detection Bias, which encompasses randomization for evaluating outcomes and the blinding of those assessing results; 4) Attrition Bias, stemming from incomplete outcome data; 5) Selective Reporting Bias; and 6) other biases, totaling ten distinct categories. If ambiguities arise, they were addressed by engaging in deliberation with an independent third-party investigator.

The evaluation was conducted utilizing STATA 18.0. Dichotomous variables were analyzed for intervention effects using the risk ratio (RR) and its 95% confidence interval (CI). For continuous variables, we measured the standardized mean difference (SMD) and their corresponding 95% CI. Statistical significance was established at P-value <0.05. Heterogeneity across study outcomes was quantitatively assessed using the χ² test in conjunction with the I² test. For study outcomes with low heterogeneity (I² ≤ 50%), a fixed-effects model was used, whereas a random-effects model was utilized for those with substantial heterogeneity (I² > 50%). To ensure the reliability of the outcomes, we carried out sensitivity analyses on indicators with significant heterogeneity (I² > 50%) by excluding studies individually to assess their impact on the combined effect.

The primary outcome indicators were analyzed through subgroup evaluations to recognize potential factors contributing to heterogeneity among the studies included. The evaluations were categorized by modeling techniques, publication periods (pre-2017 and post-2017), animal species (mice versus rats), drug dosages (≤200 mg/kg and >200 mg/kg), and treatment duration (>8 weeks and ≤8 weeks). A P-value <0.05 established statistical significance.

STATA 18.0 software was used to draw funnel plots. To assess publication bias, we examined the symmetry of the funnel plots as a primary indicator. Additionally, the Egger test can identify potential publication bias, with significance at P < 0.05.

We discovered 2,564 articles of potential relevance from ten databases. After removing 1,370 duplicate records, 1,194 articles remained for additional screening. Subsequently, 1,168 articles were excluded through title/abstract evaluation and full-text review according to predefined exclusion criteria. Ultimately, We incorporated 26 eligible studies covering the period from 2010 to 2024, reflecting a growing research interest in BBR’s therapeutic efficacy against renal fibrosis in recent years. Figure 1 illustrates the detailed process of selection.

This study encompassed 26 animal experiments (2010–2024) involving 472 renal fibrosis model animals, split evenly between a treatment group of 236 and a control group of 236. The experiments were conducted using mice and rats as the animal models, such as Sprague Dawley rats in 11 studies (42.3%), C57BL/6 mice in 7 studies (26.92%), Wistar rats in 5 studies (19.23%), and one study each using KKAy mice (3.85%), db/db mice (3.85%) and Albino rats (3.85%). Sex distribution was dominated by male animals (88.5%); 2 studies (7.7%) used female animals, and only 1 study (3.8%) did not specify sex. Most studies reported the initial weight of the animals, although five studies did not specify a weight range. Regarding model construction, the diabetic nephropathy (DN) model was the predominant induction type, accounting for 17 studies (69.2%). Among these, 14 models were induced using streptozotocin (STZ), either alone or in combination with a high-fat/high-sugar diet, while one model used alloxan. The remaining models included unilateral ureteral obstruction (UUO, 4 studies), unilateral renal artery stenosis (1 study), renal ischemia-reperfusion (I/R, 1 study), and drug-induced models (e.g., doxorubicin, cisplatin, adenine, totaling 3 studies). The dosage of BBR administered ranged from 25 to 400 mg/kg, with treatment durations ranging from 2 to 20 weeks. Regarding outcome measures, Scr levels were reported in 25 studies; BUN levels in 22 studies; TGF-β1 levels in 12 studies; α-SMA levels in 8 studies; 24-h urinary protein levels in 10 studies; KWI in 14 studies; renal fibrosis area in 8 studies; Collagen I in 5 studies; Collagen IV levels in 6 studies; FN levels in 8 studies; E-cadherin levels in 4 studies; KIM-1 levels in 3 studies; and renal injury scores in 5 studies. Several studies also measured inflammatory markers, like NLRP3 and IL-1β, along with oxidative stress parameters, including GSH-Px, MDA, and SOD. Renal histopathological assessment was dominated by hematoxylin and eosin (H&E), Masson trichrome, and Periodic Acid-Schiff (PAS) staining. Table 1 details the included studies’ characteristics.

We evaluated the methodological quality of every included study based on the specified assessment criteria. Among the 26 incorporated studies, four studies (Ma et al., 2016; Sun et al., 2020; Ma et al., 2022; Ma et al., 2024) reported specific randomization; eighteen studies (Liu et al., 2010; Huang et al., 2012; Liu et al., 2012; Xie et al., 2013; Wang et al., 2014; Miao and Zhang, 2014; Sun et al., 2015; Zhang et al., 2016; Qiu et al., 2017; Yang et al., 2017; Li and Zhang, 2017; Yu et al., 2020; Xiao et al., 2021; Ibrahim Fouad and Ahmed, 2021; Tan et al., 2023; Al-Jebouri and Al-Murshidi, 2023; Liu et al., 2023; Li et al., 2024) mentioned random allocation without specifying the randomization method, while the remaining four studies (Ni et al., 2015; Ahmedy et al., 2022; Wang et al., 2023; Cai et al., 2024) lacked information on whether they were randomly grouped. Only one study (Tan et al., 2023) reported allocation concealment, while the rest did not. Twelve studies (Liu et al., 2012; Xie et al., 2013; Ni et al., 2015; Sun et al., 2015; Ma et al., 2016; Qiu et al., 2017; Li and Zhang, 2017; Xiao et al., 2021; Ibrahim Fouad and Ahmed, 2021; Ma et al., 2022; Ahmedy et al., 2022; Ma et al., 2024) described identical animal housing conditions; fourteen studies (Liu et al., 2010; Huang et al., 2012; Wang et al., 2014; Miao and Zhang, 2014; Zhang et al., 2016; Yang et al., 2017; Sun et al., 2020; Yu et al., 2020; Tan et al., 2023; Al-Jebouri and Al-Murshidi, 2023; Wang et al., 2023; Liu et al., 2023; Cai et al., 2024; Li et al., 2024) did not provide specifics. Six studies (Liu et al., 2012; Qiu et al., 2017; Yang et al., 2017; Wang et al., 2023; Cai et al., 2024; Li et al., 2024) failed to report complete data, while the other twenty studies reported complete data. None of the studies mentioned intervention blinding, randomization for outcome assessment, or outcome blinding. Each study described how they balanced intergroup baseline characteristics. There were a lack of selective reporting bias and other bias in all studies. Methodological quality evaluation can be found in Supplementary Table S2.

In response to the elevated heterogeneity identified in the included animal experiments, a comprehensive subgroup evaluation was carried out to uncover possible causes of variation. The primary outcome indicators were stratified by publication year, treatment duration, modeling method, animal species, and drug dosages. Results suggest that drug dosages could be the source of heterogeneity for Scr and α-SMA outcomes. Notably in Scr analysis, subgroups receiving ≥200 mg/kg and <200 mg/kg doses both demonstrated reduced heterogeneity. For α-SMA outcomes, the ≥200 mg/kg subgroup exhibited more pronounced heterogeneity reduction than lower dosage groups. Regarding BUN and TGF-β1 outcomes, modeling method appeared to be the predominant source of heterogeneity. Stratification by modeling method revealed substantial heterogeneity reduction within the UUO model subgroup while other classification factors demonstrated limited impact. This result remained consistent across both BUN and TGF-β1 analyses. Supplementary Table S3 presents the detailed result.

Using the “leave-one-out” method, each outcome indicator was evaluated by excluding individual studies one at a time to evaluate the reliability of the results. The GSH-Px assessment, with the exclusion of the Wang et al. (2014) research, modified the overall effect size, yet the direction of the effect for other indicators was unchanged. Supplementary Appendix Figure S1 depicts the results.

For evaluating publication bias for Scr, BUN, TGF-β1, and 24-h urinary protein, we employed funnel plot analysis and the Egger test. Regarding 24-h urinary protein, the Egger test did not demonstrate statistical significance (P = 0.689), and the corresponding funnel plot exhibited symmetry, indicating minimal bias likelihood. In contrast, for these remaining indicators (Scr, BUN, and TGF-β1), the Egger test all demonstrated P = 0.000 < 0.05, and the corresponding funnel plots showing significant asymmetry. These findings suggest potential publication bias or lower-quality studies due to unpublished negative results or methodological heterogeneity. Supplementary Appendix Figure S2 shows the results.

Scr and BUN are essential biomarkers for evaluating renal impairment, reflecting glomerular filtration rate and nitrogen metabolism abnormalities (Lopez-Giacoman and Madero, 2015). The 24-h urinary protein level serves as an essential clinical indicator for evaluating glomerular filtration barrier’s structural integrity, with its elevated levels strongly linked to increased glomerular basement membrane permeability and podocyte damage (Zeng et al., 2024). TGF-β1, a key regulator of the pro-fibrotic process, promotes fibroblast proliferation in the renal interstitium and extracellular matrix deposition through Smad pathway activation (Ma and Meng, 2019). α-SMA serves as a definitive indicator of myofibroblast activation, with its expression intensities positively correlating significantly with the severity of renal tubulointerstitial fibrosis (Jercan et al., 2012). Considering the key pathological indicators mentioned above, we systematically explored the time-dose response characteristics of BBR in treating renal fibrosis. Our results revealed that 50 mg/kg was the minimum effective dose for improving Scr and BUN, whereas the highest administered concentration reached 400 mg/kg. The shortest effective treatment period for Scr was 2 weeks, with the longest being 20 weeks, while for BUN, it varied between 2 and 16 weeks. The therapeutic dosage range for α-SMA and TGF-β1 was also 50–400 mg/kg, with α-SMA showing an effective treatment period of 2 weeks at the shortest and 16 weeks at the longest, and TGF-β1 being effective from 2 to 12 weeks. The effective dose for 24-h urinary protein varied between 100 mg/kg and 400 mg/kg, with an effective treatment duration of 5–16 weeks. In conclusion, BBR can effectively improve renal function, reduce proteinuria, and inhibit fibrosis with optimal efficacy achieved within a dose range of 100–400 mg/kg for 5–12 weeks of treatment. These findings are illustrated in Figure 13.

We systematically assessed the time-dose response associations and the specificity of action mechanisms of berberine (BBR) across diverse renal fibrosis models by synthesizing data from 26 pre-clinical studies. The meta-analysis revealed that BBR significantly improved renal function, reduced proteinuria, mitigated renal injury, inhibited fibrosis, and attenuated oxidative stress and inflammation. In comparison to previous studies that primarily focused on single models, our multi-model synthesis offers a more comprehensive perspective on BBR’s efficacy across different fibrotic conditions. This approach enhances the understanding of BBR’s multifaceted mechanisms, supporting its broader therapeutic potential. Unlike current therapies, BBR’s multi-targeted suppression of TGF-β1/Smad signaling, oxidative stress, and inflammatory mediators may address the multifactorial nature of fibrosis, potentially overcoming therapeutic resistance in monotherapy approaches.

The time-dose-response analysis indicated that the optimal efficacy of BBR was observed at doses between 100–400 mg/kg and over a therapeutic period of 5–12 weeks. Subgroup analyses of primary outcome metrics suggested that different model types and administered doses were the main sources of inter-study heterogeneity. These findings align with prior research, which indicates that variability in fibrotic responses can result from differing experimental models and compound administration (Wynn, 2008). Additionally, another study have emphasized the importance of considering model specific responses to optimize treatment methods (Rockey et al., 2015). These insights underscore the necessity of customizing BBR applications to meet specific experimental needs. Furthermore, sensitivity analyses demonstrated that the direction of effect of most outcome indicators was insensitive to single-study exclusion, supporting the robustness of the results, except for the GSH-Px indicator, where the effect size was significantly altered after the exclusion (Wang et al., 2014) of this study. However, the publication bias assessment revealed significant asymmetry in the funnel plots for Scr, BUN, and TGF-β1, suggesting potential unpublished negative results or methodological heterogeneity, while the risk of publication bias was low for 24 h urine protein. Taken together, these findings indicate that while BBR shows significant potential for anti-renal fibrosis across different models, the small sample size and high heterogeneity highlight the need for further studies to validate the clinical relevance of these heterogeneous metrics and to refine experimental designs to minimize bias.

This research presents a comprehensive assessment of the efficacy and underlying mechanisms of BBR in combating renal fibrosis by integrating multi-model data. However, several limitations must be acknowledged. 1) the restricted quantity of studies incorporated and their heterogeneous methodological quality—particularly due to insufficient detailed information regarding randomization methods, allocation concealment, and blinding details—may introduce selection and performance biases, thus compromising the reliability of the results. 2) the high inter-study heterogeneity restricts the generalizability of the efficacy assessments. This heterogeneity primarily stems from drug dosages and modeling method. Therefore, efficacy extrapolation to lower-dose regimens or different disease models requires caution. 3) funnel plot asymmetry for key indicators suggests potential publication bias, as unpublished data or the lack of negative results may overestimate the therapeutic effects of BBR. 4) clinical translational evidence is scarce because current conclusions are based entirely on animal studies. This necessitates further validation of BBR’s effectiveness and safety through clinical trials.

Future research should prioritize integrating multi-omics technologies, such as metabolomics and single-cell sequencing, to comprehensively investigate BBR’s target networks across various pathological stages of renal fibrosis and to explore its interaction with the gut microbiota. Emerging evidence (Xu et al., 2022) suggests that changes in specific microbial taxa (e.g., Bifidobacterium, Lactobacillus) are closely linked to kidney disease progression. Furthermore, BBR has shown potential to ameliorate metabolic disorders and mitigate diverse pathologies through gut microbiota modulation (Wang et al., 2022; Yang et al., 2023). All those findings indicate that BBR may modulate renal fibrosis-related inflammation and metabolic imbalances through metabolites produced by the microbiota. Subsequent studies could establish a “BBR-microbiota-renal fibrosis” tripartite causal model to clarify its regulatory pathways along the gut-kidney axis. Furthermore, current research predominantly focuses on diabetic nephropathy models. Future studies should broaden their scope to include non-diabetic renal fibrosis models (e.g., UUO, ischemia-reperfusion) to validate the generalizability of BBR’s anti-fibrotic mechanisms and provide comprehensive experimental support for precision therapeutic strategies.