This meta-analysis followed the PRISMA (preferred reporting item for systematic evaluation and meta-analysis) guidelines (Hutton et al., 2015) and was registered with PROSPERO (NO. CRD42023492715).

To investigate the treatment of CHF with XMLI, the two reviewers conducted an extensive literature search. Relevant studies were searched and retrieved from various databases, including PubMed, Embase, Web of Science, Cochrane Library, Wanfang Data, China Knowledge Infrastructure Database (CNKI), China Biomedical Database (CBM), and China Science and Technology Journal Database (VIP). The search employed a combination of MeSH terminology and textual terminology. The search terms include “Xinmailong injection”, “Xinmailong ", “heart failure”, and “chronic heart failure”. The search was conducted from their establishment to 31 December 2023. In addition, the reviewers manually searched the reference lists of published literature to ensure comprehensive coverage. Detailed search strategies can be found in the Supplementary Material.

According to the PICOS principle, the following conditions must be met for inclusion in the study: 1) RCTs without any language restrictions on publication. 2) Participants diagnosed with CHF, aged 18 and above. 3) The intervention group received a combination of XMLI and conventional treatment (CT), while the control group received CT based on heart failure (HF) guidelines. 4) The primary outcome measures primarily focus on LVR (LVEF, LVEDD, LVESD) and inflammatory mediators (TNF-α, IL-6, IL-10, IL-18, CRP, hs-CRP).

The exclusion criteria are as follows: 1) Non-RCTs. 2) Unstable heart failure. 3) Repeated publication, retaining only complete data for research. 4) Research without primary outcome measures. 5) The full study cannot be obtained through databases or other means.

Two reviewers (XH and XC) independently evaluated the included studies and extracted data. If any discrepancies or disagreements arose during the evaluation process, a third reviewer (MY) was available for discussion and resolution. The data extraction was conducted by the two researchers (XH and XC) using a pre-established table that included several important parameters. These parameters encompassed the article title, first author, publication year, sample size, intervention drugs, dosage and course of treatment, outcome measures, and adverse reactions.

Two reviewers (YL and JY) independently evaluated the methodological quality of the included studies using the Cochrane Collaboration risk of bias tool (Sterne et al., 2019). The evaluation encompassed various aspects, including randomization methods, allocation concealment, blinding, completeness of outcome data, selective reporting, and other sources of bias. The results of the evaluations were then cross-checked to ensure accuracy and consistency. The risk of bias for each study was classified as low, unclear, or high. Any disagreements that arose during the methodological quality assessment process were resolved through discussions involving third reviewer (XM).

All meta-analyses were conducted using RevMan5.4 and Stata 17.0 software. For dichotomous data, a 95% confidence interval (CI) risk ratio (RR) was calculated, while continuous data utilized a mean difference (MD) with a 95% CI. Heterogeneity among the included studies was evaluated using I². I² ≤ 50% was considered as low heterogeneity, and a fixed-effects model was applied. Conversely, a random-effects model was applied. Furthermore, subgroup analysis was performed based on differences in LVEF to investigate possible factors influencing the results. Sensitivity analysis was performed on the primary outcome measures to evaluate the impact of individual studies on the combined effect size. The Egge’s tests were employed to test for potential publication bias.

A total of 1,402 related studies were retrieved through a systematic search. After deduplication and screening, 32 studies (Bao, 2017; Chen, 2012; Guo et al., 2016; Han, 2019; Huang and Cheng, 2022; Hui and Wu, 2017; Li et al., 2022; Li et al., 2016; Li et al., 2020; Liu et al., 2015; Mao, 2019; Song, 2022; Song et al., 2016; Su, 2020; Wang, 2019; Wei, 2023; Wu et al., 2017; Xi et al., 2019; Xu et al., 2018; Xu and Cao, 2019; Yan et al., 2019; Yao and Yang, 2022; Ye et al., 2017; Yu and Zhang, 2023; Yuan, 2019; Zhang et al., 2021; Zhang, 2020; Zhang et al., 2013; Zhao et al., 2019; Zhao, 2019; Zhou and Xu, 2020; Zhu et al., 2017) published from 2012 to 2023 were selected for the final analysis. The literature search results are displayed in Figure 1. These studies were conducted in China and involved 3,346 patients (1855 males and 1,491 females) with varying sample sizes (ranging from 23 to 175) and treatment courses (ranging from 5 to 28 days). The control group received CT recommended by the HF guidelines, while the treatment group received XMLI combined with CT. No statistically significant differences in general information were found between the two groups. The included studies provided results on various parameters, including LVEF (Chen, 2012; Zhang et al., 2013; Liu et al., 2015; Guo et al., 2016; Li et al., 2016; Song et al., 2016; Bao, 2017; Hui and Wu, 2017; Wu et al., 2017; Ye et al., 2017; Zhu et al., 2017; Xu et al., 2018; Han, 2019; Mao, 2019; Song, 2022; Wang, 2019; Xi et al., 2019; Xu and Cao, 2019; Yan et al., 2019; Yuan, 2019; Zhao, 2019; Zhao et al., 2019; Li et al., 2020; Su, 2020; Zhang, 2020; Zhou and Xu, 2020; Zhang et al., 2021; Huang and Cheng, 2022; Li et al., 2022; Yao and Yang, 2022; Wei, 2023; Yu and Zhang, 2023), LVEDD (Chen, 2012; Liu et al., 2015; Hui and Wu, 2017; Ye et al., 2017; Zhu et al., 2017; Han, 2019; Mao, 2019; Song, 2022; Xi et al., 2019; Zhao, 2019; Zhao et al., 2019; Li et al., 2020; Su, 2020; Zhou and Xu, 2020; Zhang et al., 2021; Huang and Cheng, 2022; Li et al., 2022; Wei, 2023; Yu and Zhang, 2023), LVESD (Liu et al., 2015; Song, 2022; Ye et al., 2017; Zhu et al., 2017; Zhao, 2019; Huang and Cheng, 2022; Li et al., 2022; Wei, 2023; Yu and Zhang, 2023), TNF-α (Zhang et al., 2013; Ye et al., 2017; Zhu et al., 2017; Yuan, 2019; Zhao et al., 2019; Huang and Cheng, 2022; Song, 2022; Yao and Yang, 2022; Wei, 2023), IL-6 (Zhang et al., 2013; Li et al., 2016; Song, 2022; Bao, 2017; Ye et al., 2017; Zhu et al., 2017; Yuan, 2019; Zhao et al., 2019; Huang and Cheng, 2022; Yao and Yang, 2022; Wei, 2023), IL-10 (Guo et al., 2016; Han, 2019; Zhao et al., 2019; Zhou and Xu, 2020; Wei, 2023), IL-18 (Guo et al., 2016; Han, 2019; Zhao et al., 2019; Zhou and Xu, 2020; Wei, 2023), CRP (Chen, 2012; Zhang et al., 2013; Hui and Wu, 2017; Ye et al., 2017; Zhu et al., 2017; Xu et al., 2018; Zhao, 2019), hs-CRP (Li et al., 2016; Song et al., 2016; Bao, 2017; Wu et al., 2017; Mao, 2019; Xi et al., 2019; Xu and Cao, 2019; Yan et al., 2019; Zhao, 2019; Zhao et al., 2019; Li et al., 2020; Zhang, 2020; Zhou and Xu, 2020; Zhang et al., 2021; Yu and Zhang, 2023), clinical efficacy (Chen, 2012; Li et al., 2016; Song et al., 2016; Bao, 2017; Hui and Wu, 2017; Ye et al., 2017; Zhu et al., 2017; Xu et al., 2018; Wang, 2019; Xi et al., 2019; Xu and Cao, 2019; Yuan, 2019; Zhao, 2019; Zhao et al., 2019; Li et al., 2020; Su, 2020; Zhang, 2020; Zhou and Xu, 2020; Zhang et al., 2021; Li et al., 2022; Yu and Zhang, 2023), 6-MWD (Zhang et al., 2013; Li et al., 2016; Bao, 2017; Hui and Wu, 2017; Wu et al., 2017; Xu et al., 2018; Mao, 2019; Wang, 2019; Wei, 2023), BNP (Liu et al., 2015; Song, 2022; Guo et al., 2016; Li et al., 2016; Song et al., 2016; Wu et al., 2017; Zhu et al., 2017; Xu et al., 2018; Wang, 2019; Yan et al., 2019; Zhao, 2019; Su, 2020; Zhang et al., 2021; Li et al., 2022; Yao and Yang, 2022), and NT-pro BNP (Bao, 2017; Hui and Wu, 2017; Ye et al., 2017; Mao, 2019; Xi et al., 2019; Xu and Cao, 2019; Li et al., 2020; Zhang, 2020; Zhou and Xu, 2020; Huang and Cheng, 2022; Wei, 2023; Yu and Zhang, 2023). Among these studies, only 10 studies (Ye et al., 2017; Zhu et al., 2017; Xu et al., 2018; Xi et al., 2019; Li et al., 2020; Su, 2020; Zhang et al., 2021; Huang and Cheng, 2022; Li et al., 2022; Yao and Yang, 2022) reported adverse reactions. The basic characteristics of the included studies are present in Table 1.

The overall quality of the included studies varied. 20 studies (Bao, 2017; Huang and Cheng, 2022; T. T; Li et al., 2016; Li et al., 2020; Mao, 2019; Song, 2022; Su, 2020; Wei, 2023; Wu et al., 2017; Xu et al., 2018; Yan et al., 2019; Yao and Yang, 2022; Ye et al., 2017; Yu and Zhang, 2023; Yuan, 2019; Zhang et al., 2021; Zhang, 2020; Zhao et al., 2019; Zhao, 2019; Zhou and Xu, 2020) utilized the low-risk random number table method. Conversely, 10 studies (Chen, 2012; Zhang et al., 2013; Liu et al., 2015; Guo et al., 2016; Song et al., 2016; Hui and Wu, 2017; Wang, 2019; Xi et al., 2019; Xu and Cao, 2019; Li et al., 2022) lacked clear descriptions of randomization, resulting in an unclear risk assessment. Two studies (Zhu et al., 2017; Han, 2019) were considered high-risk as they grouped patients based on admission order. None of the studies reported hidden allocation, leading to an unclear risk assessment for all of them. In terms of design, four studies (T. T. Li et al., 2016; Mao, 2019; Yan et al., 2019; Zhao, 2019) were multicenter double-blind tests, which were considered to be low-risk. Additionally, all included studies were published during a period of low risk of selective reporting and were given priority based on their locality. However, none of the studies clearly indicated the presence of other biases, resulting in an overall unclear risk assessment. The risk of bias assessment is detailed in Figure 2.

To assess the reliability and robustness of the consolidation results, a sensitivity analysis was conducted. This analysis involved sequentially excluding individual studies and examining their impact on various variables, including LVEF (Figure 14A), LVEDD (Figure 14B), LVESD (Figure 14C), TNF-α (Figure 14D), IL-6 (Figure 14E), IL-10 (Figure 14F), IL-18 (Figure 14G), CRP (Figure 14H), and hs-CRP (Figure 14I). Interestingly, the exclusion of any of these studies had no significant effect on the combined results. This finding suggests that the merged results are both robust and reliable, as clearly shown in Figure 14.

In order to evaluate publication bias, the Egger’s test was utilized specifically for LVEF, LVEDD, IL-6, hs-CRP, BNP, and NT-pro BNP, as depicted in Figure 15. Remarkably, the results of the analysis revealed that there was no significant publication bias for LVEF (Figure 15A, p = 0.667), LVEDD (Figure 15B, p = 0.188), IL-6 (Figure 15C, p = 0.500), hs-CRP (Figure 15D, p = 0.836), BNP (Figure 15E, p = 0.767), and NT-pro BNP (Figure 15F, p = 0.298).

This meta-analysis is the first to investigate the effects of XMLI on LVR and inflammatory mediators in patients with CHF. A total of 32 RCTs were included in this analysis, revealing several important findings. Firstly, the combination of XMLI and CT significantly improved LVR in HF patients. This improvement was supported by an increase in LVEF, as well as a decrease in LVEDD and LVESD. As well as levels of BNP, and NT-pro BNP were decreased. Furthermore, the combination therapy also resulted in a significant reduction in inflammatory mediators. Specifically, there was a decrease in the expression levels of pro-inflammatory cytokines TNF-a, IL-6, IL-18, CRP, and hs-CRP. Conversely, there was an increase in the expression levels of the anti-inflammatory cytokine IL-10. In addition to improving LVR and reducing inflammatory mediators, the combination therapy showed higher clinical efficacy and improvement of the 6-MWD. Importantly, the combination therapy demonstrated good safety, with only minor adverse events reported. These events were manageable in terms of symptoms and had no impact on treatment outcomes. Based on these compelling results, our meta-analysis suggests that XMLI effectively improves LVR in HF patients, reduces inflammatory mediators, and enhances overall clinical efficacy.

To ensure the robustness and reliability of our findings, sensitivity analysis was conducted. Individual studies were sequentially deleted, and sensitivity analysis was performed on key indicators of LVR and inflammatory mediators, further confirming the validity of our results. Additionally, Egger’s test was conducted to evaluate publication bias, with the results showing no significant publication bias. This further strengthens the validity and reliability of our findings.

Although Lu et al.'s (Lu et al., 2018) previous meta-analysis evaluated the clinical efficacy of XMLI in treating CHF, our study specifically focuses on its effects on LVR and inflammatory mediators. It should be noted that there are several shortcomings in previous studies: Firstly, HF is classified into different subtypes based on ejection fraction according to the HF management guidelines. However, previous studies did not consider these subtypes or conduct subgroup analysis based on ejection fraction during meta-analysis, which may introduce heterogeneity and bias. Secondly, sensitivity analysis was not performed, and only funnel plots were used to evaluate publication bias in previous studies, which affects the robustness and reliability of the research findings. Finally, previous studies mainly focused on clinical indicators such as symptom improvement, exercise tolerance, and quality of life in CHF patients treated with XMLI. The impact of XMLI on LVR and inflammatory mediators has not been extensively explored. Understanding these specific effects is crucial for a comprehensive assessment of XMLI’s therapeutic potential in CHF patients.

This meta-analysis of RCTs is the first to specifically investigate the effects of XMLI on LVR and inflammatory mediators in patients with CHF. To enhance the reliability of our findings, we will conduct subgroup analyses based on different types of HF, eliminating potential confounding factors associated with disease types. Additionally, our study addresses a crucial aspect of CHF by evaluating the impact of XMLI on LVR, a key pathological and physiological mechanism contributing to high hospitalization and mortality rates in CHF patients. Furthermore, we comprehensively evaluate the role of inflammatory response in LVR, highlighting its significance in the progression of HF, an aspect that previous studies have yet to fully address.

Several limitations should be considered when interpreting the results of this study. Firstly, the included studies in this meta-analysis exhibit relatively low overall quality, with limited reporting of allocation concealment and blinding, which poses a serious risk of bias. Secondly, significant heterogeneity is observed among the RCTs, although subgroup and sensitivity analyses are conducted without identifying the sources of heterogeneity. This variation may be associated with the lack of standardized dosage and intervention duration. Thirdly, the small size of the included studies highlights the need for larger scale research to ensure result reliability. Therefore, caution should be exercised when interpreting the findings. Fourthly, it is important to note that all studies included in this analysis were conducted in China and exclusively involved Chinese participants. This limited geographical scope may introduce sources of heterogeneity. To ensure the applicability of these findings to different races, future studies should incorporate with more diverse samples from various geographical regions. Lastly, the limited number of studies examining TNF-α, IL-10, IL-8, and CRP results in low supporting evidence. Future research should prioritize larger scale and more rigorous studies to verify the stability of our findings.

To strengthen the evidence regarding the efficacy of XMLI treatment for CHF, future clinical research should address the following areas. Firstly, studies should be conducted on different types of HF to comprehensively assess the efficacy of XMLI in treating HF. By doing so, bias can be minimized, and more accurate conclusions can be generated. Secondly, rigorous adherence to clinical research standards, including strict randomization, allocation concealment, and blinding techniques, should be ensured. Encouraging placebo-controlled randomized trials can provide more precise results. Thirdly, it is crucial to report RCTs in a complete and comprehensive manner by employing standardized reporting trial statements. Alongside primary outcome measures, additional information such as comorbidities, disease duration, medication usage, readmission rates, follow-up duration, and endpoint time should be reported, facilitating the analysis of heterogeneity and prognosis clarification. Finally, given the significant role of LVR and inflammation in the progression of HF, future research should focus on targeted and high-quality studies in these areas.