We conducted this report based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses checklist. This study was registered with PROSPERO (registration number: CRD42023402908). We searched the following electronic bibliographic databases: Web of Science, PubMed, Cochrane Library, and CNKI(China National Knowledge Infrastructure). Studies that reported the efficacy of TCM in animals with DCM were included. The literature search was conducted using the terms (“Diabetic cardiopathy”) OR (“diabetic cardiomyopathy”)OR (“Autonomic neuropathy") OR (“microangiopathy”) OR (“microvascular disease”) AND (“animal” OR" animals” OR “rat” OR” rats” OR” mouse”OR” mice”) AND (“Chinese medicine “OR” Chinese herb ” OR” herbal ” OR ” natural drug” OR “Natural Product” OR “traditional Chinese” OR “formula”). The date will be restricted from the year 2013 to 24 April 2023. The English and Chinese languages will be searched only. All searches were performed by two independent researchers (LH and LQ).

The criteria for inclusion in this study consist of animal models with diabetic cardiomyopathy without limitations regarding the construction, design types, randomized controlled trials, sexes and species. All studies that comprise the number of animals used and TCM treatments, including a single Chinese herbal extract (monomer), compounds of several Chinese herbal extracts, and Chinese herbal compounds, were included. Controlled studies with one or more separate control groups and those published in Chinese or English were also included. On the other hand, the exclusion criteria consisted of cell and in vitro studies, studies that used other non-TCM treatments, case reports, clinical trials, reviews, and cross-over studies that lacked a separate control group. Furthermore, studies that did not provide data were not included. Two researchers (AS and JR) independently screened the studies, and discrepancies were resolved through discussions.

Data extraction from the included studies was conducted by two independent researchers (LH and LQ). The following data were extracted from the studies: outcome measures (body weight, HW/BW, blood glucose, triglyceride, serum total cholesterol, left ventricular end-systolic and end-diastolic diameter, ejection fraction, fractional shortening, and E/A ratio), publication information (author and year), details of treatments administered (dose, route, and duration), and the species of animals used. The study extracted sample size, mean value, and standard deviation (SD)/standard error for the both control and treatment groups for each comparison. In cases where multiple treatment groups shared the same control group, the sample size of the control group was divided by the number of treatment groups for appropriate comparison (Eriksson and Nyström, 2015). If outcomes were measured at more than one time point, we only included data from the last time point (Xiao and Luo, 2018).

The included studies were independently assessed for quality by two researchers, and disagreements were resolved through consultation with a third party. The quality of the included studies was evaluated using the SYRCLE (Systematic Review Center for Laboratory Animal Experimentation) Risk of Bias tool. The assessment includes 11 areas that are scored and evaluated separately. These areas are as follows: 1) Random sequence generation; 2) Baseline characteristics; 3) Allocation concealment; 4) Random housing; 5) Blinding; 6) Blinding of participants and personnel; 7) Random outcome assessment; 8) Blinding of outcome assessment; 9) Incomplete outcome data; 10) Selective reporting; 11) Other bias. Any discrepancies that arose during the evaluation were resolved through discussion until a consensus was reached.

Data were synthesized using either a fixed effects model or a random effects model based on a heterogeneity test. Heterogeneity between studies was explored by Cochran’s Q statistic and I²statistic (Diao et al., 2019). All data were analyzed using Review Manager version 5.4 and STATA version 17 packages. Standardized mean differences (SMD) were used as an index of effect for each effect, their point estimates, and 95% confidence intervals. Heterogeneity between included studies was analyzed using the χ² test, while the magnitude of heterogeneity was determined quantitatively using I². If heterogeneity between results was not significant (I² ≤ 50%), meta-analysis was performed using a fixed effects model (EF, FS, and E/A). If the heterogeneity between studies was significant (I²>50%), a random effects model was used (body weight, HW/BW, BG, TG, TC, and LVIDs). The study’s findings were mainly presented using forest plots, with funnel plots utilized to analyze publication bias. Additionally, sensitivity analyses and Egger’s test were carried out (Feidantsis et al., 2018). To further explore the sources of heterogeneity, subgroup analyses should be conducted on the types of Traditional Chinese Medicine (TCM), duration, modeling method, and animal species.

In this study, a search of 199 publications was conducted, with 168 articles obtained from PubMed, 19 from CNKI, 2 from Web of Science, and 10 from manual search. The duplicates were screened using Endnote 20, resulting in 91 exclusions. Based on the title and abstract assessment, 76 articles were excluded manually, and 8 out of the remaining 32 articles were excluded after reading the full text for not meeting the inclusion criteria. Eventually, a total of 24 articles were included in this study. Please refer to Figure 2 for the screening process details (Page et al., 2021). Altogether 2157 animals (1088 in the TCM group; 1069 in the control group) were included in this meta-analysis. Ten studies used Sprague Dawley rats, seven studies used Wistar rats, six studies used C57BL/6 J mice, and one study used Kunming mice. Rats weighed between 70–280 g, while C57BL/6 J mice and Kunming mice weighed between 23-27.5 and 18–22 g, respectively. The age of the rats was only reported in ten studies and ranged from 6 to 12 weeks. The studies used three different models: High-fat food (HFD) and streptozotocin (STZ) in eleven studies, STZ alone in ten studies, and alloxan in three studies.

Regarding the intervention methods, the researchers identified 24 different types of TCM, including ten compounds and fourteen monomers. The monomers consisted of Berberine (An isoquinoline alkaloid, mainly derived from Rhizoma Coptidis [Ranunculaceae; Coptis chinensis Franch. rhizome]), Curcumin (A yellow curry pigment isolated from Turmeric [Zingiberaceae; Curcuma longa L. root]), Crocin (An extract of Crocus sativus [Iridaceae; Crocus sativus L. flower]), Dioscin (An extract of Dioscoreaceae [Dioscoreaceae; Dioscorea oppositifolia L. rhizome]), Mangiferin (A carboketone glycoside of tetrahydroxypyridone from rhizoma anemarrhenae [Asparagaceae; Anemarrhena asphodeloides Bunge root]), Resveratrol (A non-flavonoid polyphenolic organic compound isolated from the roots of Veratrum grandiflorum [Melanthiaceae; Veratrum grandiflorum root]), Myricetin (A flavonol compound extracted from the bark of Myrica rubra [Myricaceae; Myrica rubra]), Nar (A dihydroflavonoid from pomelo [Rutaceae; Citrus paradisi Macfadyen L. fruit]), Polydatin (An extract of Plantago Ovata [Plantaginaceae; Plantago ovata Forssk. rhizome]), Rutin (A natural flavonoid glycoside present in Sophora japonica rice [Fabaceae; Styphnolobium japonicum L. buds]), Sal B (the representative component of phenolic acids derived from the Salvia miltiorrhiza Bunge [Lamiaceae; Salviae miltiorrhizae radix et rhizoma]), taxifolin (a flavonoid commonly found in Pseudotsuga taxifolia [Pinaceae; Pseudotsuga menziesii var. menziesii]), triptolide (a most important active ingredient extracted from Tripterygium [Celastraceae, Tripterygium wilfordii Hook.f.]), and Wogonin (an important flavonoid derived from the Scutellaria [Lamiaceae; Scutellaria baicalensis Georgi root]). The intervention durations ranged from 24 to 180 days. Intraperitoneal administration was performed in two of the 24 studies, while gastrointestinal administration was used in the remaining 22 studies. The primary outcome measurements included BW (reported in 11 studies), HW/BW (reported in 7 studies), BG (reported in 15 studies), TG (reported in 7 studies), TC (reported in 10 studies), LVIDs (reported in 6 studies), EF (reported in 8 studies), FS (reported in 8 studies), and E/A (reported in 4 studies). The detailed characteristics of the included studies are shown in Table 1.

The results of the quality assessment of the 24 included studies are presented in Figure 3. None of the studies reported the random sequence generation, allocation concealment, and blinding of participants and personnel. However, all the studies reported similar baseline characteristics between the model and control groups. Additionally, all experimental animals were housed in identical conditions, so the placement of animals in all studies can be considered consistent with the principle of randomization. One study had a high risk of bias for incomplete outcome data (attrition bias) due to an unclear number of animals. Eighteen studies reported the random outcome assessment, while 12 studies assessed the blinded methods used for outcome assessment. None of the studies had a selective reporting bias or any other sources of bias. Despite the overall poor quality of these studies, none of them were excluded based on quality or risk of bias assessment.

The study employed funnel plots (Supplementary Figure S3) and Egger’s test (Supplementary Figure S4) to identify potential publication bias. The results show that there was a significant publication bias related to body weight, HW/BW, blood glucose, triglyceride, and total cholesterol (p = 0.003, p < 0.001, p = 0.001, p < 0.001 and p = 0.006 respectively; Supplementary Figure S3A–E) However, no significant publication bias was found for ejection fraction and fractional shortening (p = 0.910 and p = 0.600, respectively; Supplementary Figure S3F–G). The study did not conduct a publication bias analysis on LVIDs and E/A indicators due to a limited number of included studies.

The study conducted sensitivity analyses by individually removing each study to assess its impact on the meta-analysis results. Supplementary Figure S5 shows all effect values falling within the 95% CI of the overall meta-analysis, indicating comparable effect sizes among studies. The results suggest the statistical validity and reliability of the findings.

The results show high heterogeneity for BW, HW/BW, BG, TG, and TC. Therefore, subgroup analyses were performed to reveal the sources of heterogeneity, including variations in types of TCM, duration, modeling methods, and animal species.

The pathogenesis of diabetes encompasses various mechanisms that contribute to pancreatic beta-cell dysfunction and insulin resistance, including oxidative stress, endoplasmic reticulum stress, inflammation, autophagy, and DNA methylation (Ganesan et al., 2020). These mechanisms play crucial roles in the development of the disease. When exposed to elevated levels of glucose or fatty acids, the myocardium triggers the activation of NF-κB (Rahimi-Madiseh et al., 2017), a protein responsible for regulating immune responses and inflammation. NF-κB promotes the formation of NLRP3 inflammatory vesicles that release pro-inflammatory mediators (Asgary et al., 2014). Furthermore, heightened glucose levels disrupt mitochondrial function, leading to the generation of ROS (Hu et al., 2022). These ROS cause the aggregation of TXNIP with TRX (Feng et al., 2022), which then bind to NLRP3, initiating the activation of the NLRP3 inflammasome. This inflammasome activates caspase-1, which processes pro-inflammatory cytokines IL-1β and IL-18. ROS also stimulate NF-κB, thereby further facilitating the assembly of the NLRP3 inflammasome and the cleavage of caspase-1 (Egger et al., 1997; Engels et al., 2000). Consequently, this cascade results in the release of IL-1β and IL-18, potent inflammatory molecules implicated in the pathogenesis of diabetes. In diabetes, activated endothelial cells contribute to the early development of myocardial stiffness and diastolic dysfunction (Petersen and Shulman, 2018). They do so by promoting the uncoupling of endothelial NOS, resulting in the production of superoxide, hydrogen peroxide, and peroxynitrite. Consequently, the levels of NO are reduced (Zhou et al., 2010).

Based on this systematic review and meta-analysis, it is possible to infer that TCM has a significant impact on downregulating BG, TG, and TC in DCM animals, indicating its ability to modulate lipids and blood glucose. Moreover, the TCM treatment was highly effective in reducing myocardial hypertrophy and improving cardiac functions, as evidenced by the significant downregulation of HW/BW and the upregulation of EF, FS, and E/A with statistical significance.

Table 2 presents the outcome measures, which are divided into two subgroups based on the type of traditional Chinese medicine (TCM): monomer and compound. Based on the results of subgroup analysis, it can be inferred that herbal monomers have a more pronounced impact on BG and TG levels in animals with DCM. Moreover, the treatment of DCM is closely linked to glucolipid metabolism (Montaigne et al., 2021). Thus, these findings imply that individual herbal monomers in TCM may offer greater advantages in enhancing glucose and lipid metaboism compared to TCM compounds.

Table 3 illustrates that even after categorizing the outcome measures into two subgroups based on their duration: ≤56 days and >56 days, a significant level of heterogeneity persisted across most subgroups. This implies that duration has a negligible impact on heterogeneity.

Table 4 showcases the outcome measures, which are categorized into three subgroups based on the animal modeling methods: STZ, HFD + STZ, and alloxan. While there was still some heterogeneity present in all subgroups, reductions were observed. Therefore, we can conclude that the primary source of heterogeneity in the included studies is the modeling approach, which aligns with previous research findings (Chen et al., 2023). The index related to lipid metabolism (TG) displayed significantly better results in the STZ group compared to the HFD + STZ group, potentially due to the impact of the high-fat diet on the measurement outcomes.

Table 5 demonstrates two subgroups, mice and rats, based on animal species. Results suggest that animal species is a contributing factor to heterogeneity, contradicting the outcomes of previous meta-analyses conducted on animal models of diabetes (Chen et al., 2022). Such disparity may stem from the inclusion of different animal strains within the study.

Most early studies were superficial and primarily concentrated on blood glucose and lipids, with few studies reported on diabetic cardiomyopathy and its mechanisms. The limited literature on diabetic cardiomyopathy could potentially compromise the study’s objectivity. Secondly, the quality of the literature acquired for the study was suboptimal, particularly with respect to randomization and blinded implementation. While most studies used the term “randomization,” they did not specifically detail the randomization method used. Additionally, none of the studies mentioned the application of blinded methods in the methods section. Thirdly, the high heterogeneity observed in the forest plots is a regular issue in meta-analyses of animal research, partly due to substandard study quality. Additionally, it could stem from various experimental factors, for example, the types of TCM, modeling methods, duration, and animal species.