This systematic review and meta-analysis of RCTs were conducted by PRISMA guidelines (Moher et al., 2009). The protocol studied has been registered with PROSPERO under the registration number CRD42023453293.

A computerised search was conducted on the CNKI, WF, VIP, CBM, PubMed, Embase, Web of Science, and Cochrane Library databases for RCTs of CDDP adjuvant therapy for AMI up to May 2023. There is no restriction on the language of the search using subject terms. The search terms are Compound Danshen Dropping Pill, Danshen Dropping Pill, Acute Myocardial Infarction, Percutaneous Coronary Intervention, PCI, and AMI. Search strategies were provided in Supplementary Material S1.

Two researchers independently read the full text and extracted relevant information. The extracted information included basic information, intervention methods, risk of bias assessment, relevant outcome indicators and adverse reactions, etc. When the two parties disagreed on the inclusion of the literature, it was referred to a third party for discussion and judgment.

The Cochrane risk-of-bias tool for randomized trials was used to assess each study’s risk of bias (Sterne et al., 2019). This tool comprises a five-item checklist: 1) randomization process; 2) deviations from the intended interventions; 3) missing outcome data; 4) measurement of the outcome; and 5) selection of the reported result. The risk of each domain was examined as “low”, “high” and “some concerns”.

The included data were statistically analyzed using STATA 17.0. Continuous data was pooled with Std mean difference (SMD) and 95% confidence interval (CI), dichotomous data was pooled with relative risk (RR) and 95% CI. Heterogeneity was judged based on the results of the I² test; when I²<50%, a fixed-effects model was used, and when I² ≥ 50% indicated that inter-study heterogeneity was significant, so the reasons for heterogeneity were analyzed. First, the raw data were checked for correctness, and second, if heterogeneity was attributed to treatment duration, sample size, publication time, etc., meta-regression and subgroup analyses could be used to investigate the sources of heterogeneity. Sensitivity analysis evaluates the robustness and reliability of the results. If an outcome had more than 10 articles, a funnel plot was analysed for publication bias and publication bias was evaluated using Egger’s test with STATA 12.0.

Trial sequential analysis (TSA) was performed using TSA 0.9.5.10 Beta software (Greco and Capodanno, 2024). Evidence confidence of meta results was evaluated by GRADE (Grading of Recommendations Assessment, Development and Evaluation) according to the instructions (Brozek et al., 2009).

A total of 543 studies were searched. The retrieved titles were imported into EndNote X9, and 23 studies (Li et al., 2010; Li et al., 2011; Lin, 2011; Fang et al., 2017; Pan et al., 2017; Zhang, 2017; Li, 2018; Guihong et al., 2018; Hong et al., 2018; Ji et al., 2019; Li et al., 2019; Liu and Zhang, 2019; Shi, 2019; Su, 2019; Wenhui, 2019; Xiang, 2019; Xu et al., 2019; Zhao et al., 2019; Jian, 2020; Li, 2020; Niu, 2020; Shi, 2020; Chen, 2023) were finally included after checking and screening (Figure 1).

A total of 23 studies (Li et al., 2010; Li et al., 2011; Lin, 2011; Fang et al., 2017; Pan et al., 2017; Zhang, 2017; Li, 2018; Guihong et al., 2018; Hong et al., 2018; Ji et al., 2019; Li et al., 2019; Liu and Zhang, 2019; Shi, 2019; Su, 2019; Wenhui, 2019; Xiang, 2019; Xu et al., 2019; Zhao et al., 2019; Jian, 2020; Li, 2020; Niu, 2020; Shi, 2020; Chen, 2023) were included in this meta-analysis, all of which were conducted in China and involved 2,732 AMI patients, including 1,377 in the treatment group and 1,355 in the control group. In all studies, there were no statistical differences between the experimental and control groups in terms of age and sample size. LVEF was reported in 19 studies (Li et al., 2010; Li et al., 2011; Lin, 2011; Pan et al., 2017; Zhang, 2017; Guihong et al., 2018; Ji et al., 2019; Li et al., 2019; Liu and Zhang, 2019; Shi, 2019; Su, 2019; Wenhui, 2019; Xiang, 2019; Xu et al., 2019; Zhao et al., 2019; Jian, 2020; Li, 2020; Niu, 2020; Shi, 2020), TCER in 6 studies (Guihong et al., 2018; Li et al., 2019; Xiang, 2019; Xu et al., 2019; Zhao et al., 2019; Shi, 2020), LVEDD in 9 studies (Fang et al., 2017; Ji et al., 2019; Li et al., 2019; Wenhui, 2019; Xiang, 2019; Xu et al., 2019; Jian, 2020; Li, 2020; Niu, 2020), NT-proBNP in 3 studies (Fang et al., 2017; Li et al., 2019; Shi, 2020), cTnT in 2 studies (Ji et al., 2019; Chen, 2023), CK-MB in 3 studies (Ji et al., 2019; Li et al., 2019; Zhao et al., 2019), hs-CRP in 4 studies (Li, 2018; Shi, 2019; Zhao et al., 2019; Niu, 2020), IL-6 in 6 studies (Fang et al., 2017; Pan et al., 2017; Li, 2018; Hong et al., 2018; Zhao et al., 2019; Niu, 2020), TNF-α in 5 studies (Pan et al., 2017; Li, 2018; Hong et al., 2018; Shi, 2019; Niu, 2020), and adverse reactions/adverse events in 11 studies (Li et al., 2010; Li, 2018; Guihong et al., 2018; Ji et al., 2019; Li et al., 2019; Liu and Zhang, 2019; Wenhui, 2019; Zhao et al., 2019; Jian, 2020; Niu, 2020; Shi, 2020) (Table 1).

Note: T: Treatment group; C: Control group; ADR/AE: Adverse Reactions/Adverse Events.

In the 23 included studies, 13 used a random-number table, and the aspect of whether the order of allocation was randomized was described as “low risk”. One study was randomized using order of presentation, which may have been at risk of bias, described as “high risk”. All studies did not mention allocation hiding, so the randomization process section was listed as “some concern”. Three studies were blinded to subjects and implementers, and no information was described for the other 19 studies. There was no missing data and no selective results in any of the studies, so these sections were described as “low risk” (Figure 2).

Six studies (Guihong et al., 2018; Li et al., 2019; Xiang, 2019; Xu et al., 2019; Zhao et al., 2019; Shi, 2020) reported TCER with little heterogeneity (I² = 71.1%), therefore, meta-analysis using fixed effect model showed statistically significant between the two groups (RR = 1.17, 95% CI (1.10, 1.24), p < 0.0001), which result suggests that CDDP added to CWT is superior to CWT in improving TCER (Figure 8A). Sensitivity analysis showed that excluding any study did not affect the overall estimate of effect (Figure 8B).

Eleven studies (Li et al., 2010; Li, 2018; Guihong et al., 2018; Ji et al., 2019; Li et al., 2019; Liu and Zhang, 2019; Wenhui, 2019; Zhao et al., 2019; Jian, 2020; Niu, 2020; Shi, 2020) reported adverse reactions/adverse events with fine homogeneity, so meta-analysis using fixed-effects model showed statistically significant between the two groups (RR = 0.60, 95% CI (0.45, 0.79), p < 0.0001), suggesting that CDDP added to CWT was superior to CWT (Figure 12). The incidence of adverse reactions/adverse events was 11.6% (62/535) in the treatment group and 18.6% (96/515) in the control group.

We assessed publication bias for metrics that included more than 10 studies. The LVEF funnel plot exhibited left-right asymmetric distribution characteristics, indicating potential publication bias. The quantitative results of Egger’s test are consistent with the qualitative results of the funnel plot (t = 2.19, 95%CI, 0.096 to 4.895, p = 0.042) (Figures 13A,B). It may be related to the quality and sample size of the included studies, and selective reporting cannot be ruled out due to lack of information on clinical trial registration or study protocols. The adverse reaction/adverse event funnel plot exhibited left-right symmetry, and the quantitative results of the egger’s test were consistent with the qualitative results of the funnel plot (t = −2.13, 95%CI, −397.27 to−2.28, p = 0.048) (Figures 13C,D).

The GRADE evaluation tool was used to grade the evidence for each study’s outcome indicators, taking into account the risk of bias, inconsistency of results, directness of evidence, precision of evidence, and publication bias. Downgrading the quality of the evidence when one or more factors are present in the results of each study. TCER was evaluated as moderate. LVEF, LVEDD, IL-6, TNF-α were evaluated as low. hs-CRP, CK-MB, cTnT, and NT-proBNP were evaluated as very low because of the risk of bias associated with small sample sizes for these indicators (Table 4).

Note: ¹the included studies have certain defects in randomization, allocation concealment, and blinding; ² the included studies are highly heterogeneous; ³ Relatively few patients were included.

TSA was performed on LVEF. Type I error was defined as 5%, the information axis was set to a cumulative sample size, statistical efficacy was 80%, and the cumulative Z-value crossed the traditional and TSA bounds after item 2 to obtain a positive conclusion in advance, using the sample size as the required information size (RIS). The penalized curve exceeded the traditional bound of Z = 1.96, further confirming the clinical efficacy of CDDP in improving LVEF (Figures 14A,B).

CDDP is a highly dispersed state of Chinese patent medicine made of Salviae Miltiorrhizae Radix et Rhizoma, Notoginseng Radix, and Borneolum Syntheticum as the main ingredients and processed by modern technology, which has the effect of activating blood and activating stasis, regulating qi and relieving pain. The inflammatory response induced after PCI promotes an excessive increase in oxygen free radicals, which triggers ventricular remodeling and myocardial injury and is an initiating factor for thrombus re-formation (Li et al., 2023). Pharmacological studies have found CDDP to have coronary artery dilation, anti-inflammatory, anti-thrombotic, and vascular endothelial and cardiomyocyte protection effects, thereby further reversing IRI (Bei et al., 2017). Panax notoginseng (PNS), an active ingredient in Notoginseng Radix, can effectively inhibit oxygen sugar deprivation-induced apoptosis, probably by inhibiting oxidative stress and inducing Akt phosphorylation (Chen et al., 2011). PNS also reduces the expression of pro-inflammatory factors by inhibiting the RAGE, MAPK signaling pathway in apoE (−/−) mouse lesions (Dou et al., 2012). Ginsenosides Rb1 and Rg1 activate nitric oxide (NO) and promote endothelium-dependent vasodilatation by regulating endothelial cell PI3K/Akt/eNOS pathway and L-arginine transport (Pan et al., 2012). Ginsenoside Rd prevents the development of atherosclerosis by inhibiting Ca²⁺ inward flow through voltage-independent Ca²⁺ channels (Li et al., 2011). CDDP has a wide range of applications in cardiovascular disease, and most RCTs have demonstrated significant clinical efficacy of CDDP-assisted therapy in AMI undergoing PCI. However, trials with large sample sizes are lacking. This meta-analysis provides a valid evidence-based rationale for the clinical use of CDDP in the treatment of AMI undergoing PCI.

Meta-analysis results showed that CDDP has good positive utility for AMI undergoing PCI treatment, can reduce hs-CRP, TNF-alpha, IL-6 levels, inhibit myocardial local inflammatory response, improve myocardial injury indexes such as CK-MB, NT-proBNP, cTnT, improve LVEF, reduce LVEDD, and at the same time, it shows strong advantages in reducing the incidence of adverse reactions/adverse events. Due to the low heterogeneity, we can clarify that CDDP combined with CWT is superior to CWT in improving TCER. Interestingly despite our choice of randomized models for data statistics, sensitivity analyses, meta-regression analyses, and subgroup analyses to eliminate heterogeneity between studies, a small number of outcome metrics remained highly heterogeneous, prompting us to be cautious in interpreting the results in clinical practice. GRADE results showed that the quality of evidence ratings for most outcome indicators were very low and low. Currently, a number of studies are rated as low quality due to methodological limitations, mainly due to risk of bias, inconsistent results, indirect evidence. Most studies lacked blinding and allocation concealment designs during randomization. However, the emerging view is that unblinded pragmatic trials should be recommended because they emphasize practical applicability and extrapolation to improve the external validity of real-world trials rather than treatment effects, even though this approach is a major contributor to the low quality of included studies (Sox and Lewis, 2016). TSA was performed to exclude possible false-positive results in order to improve the robustness of the results of the meta-analysis. The results of the TSA showed that the total number of samples collected in this study met the requirements of the meta-analysis, which excluded the possibility of false-positive results, and further confirmed the efficacy of the CDDP for the treatment of AMI undergoing PCI.

The purpose of this study is to summarize and evaluate the efficacy and safety of CDDP for the treatment of AMI undergoing PCI according to the updated and optimized methods of PRISMA, so that the results are sufficiently comparable and convincing. Nevertheless, despite the encouraging results, there are inevitable limitations to this study: 1) The overall quality of the studies was poor, with most being single-center, small-sample trials and with a lack of uniformity in reporting methods for outcome indicators. The poor quality of some of the studies resulted in high risk and heterogeneity between studies, which may reduce the robustness of the results of meta-analysis. 2) As original studies related to most of the outcome metrics were scarce, only studies on LVEF improvement were assessed for publication bias. 3) In addition to the indicators analyzed in this study, there are individual studies that reported TCM efficacy scores, triglycerides (TG), and total cholesterol (TC), etc. However, due to the small number of studies and the inconsistency in the way the outcome indicators were measured, this study did not perform a combined analysis, which is to be enriched and improved in the future. 4) The evaluation of drug efficacy in the related field requires a long period of time to be completed due to the lengthy process of further treatment after PCI. The study’s inclusion of a short course of RCT and the lack of long-term follow-up means that the drug’s long-term efficacy is not yet known. 5) There were few studies that explicitly stated that they conducted their RCTs in accordance with the Consolidated Standards of Reporting Trials (CONSORT), and most did not provide registry information. 6) The lack of relevant information on blinding and allocation concealment in a small number of studies may lead to an exaggerated effect of the results.

In the future, clinical RCTs on the adjuvant treatments of CDDP against AMI should be improved in the following aspects. First, at the design stage, protocols for RCTs should comply with the CONSORT statement. High-quality clinical data conforming to international standards such as the CONSORT statement is absolutely essential for the further development and acceptance of TCM, and the improvement of the quality of clinical data will be a key step in the future of TCM. Second, RCTs should use a rigorous blinded design with recognized and uniform outcomes to evaluate the therapeutic efficacy of CDDP. Third, RCTs should perform sample size estimation and be designed with a sufficiently long follow-up period for evaluation. Fourth, improvements in TCM symptoms should be recorded in detail to provide evidence for future TCM consultations and treatments to better improve quality of life.