The protocol for this study was registered in the International Prospective Register of Systematic Reviews (ID:CRD42024603769). This study followed the reporting guideline of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 (PRISMA 2020) (Page et al., 2021).

As this is a literature review, there is no need for ethical approval.

The two researchers independently searched four Chinese databases (CNKI, Wanfang, VIP, and SinoMed) and four English databases (PubMed, Cochrane Library, Embase, and Web of Science). The search time range was from the establishment of the database to 1 June 2024. Using a combination of subject terms and free-text terms for searching. The search query in PubMed is shown below, and the search queries for other databases can be found in Supplementary Material S2.

#1 “Cerebral Hemorrhage” [Mesh] OR “Hemorrhage, Cerebrum [Title/Abstract]” OR “Cerebrum Hemorrhage [Title/Abstract]” OR “Cerebrum Hemorrhages [Title/Abstract]” OR “Hemorrhages, Cerebrum [Title/Abstract]” OR “Cerebral Parenchymal Hemorrhage [Title/Abstract]” OR “Cerebral Parenchymal Hemorrhages [Title/Abstract]” OR “Hemorrhage, Cerebral Parenchymal [Title/Abstract]” OR “Hemorrhages, Cerebral Parenchymal [Title/Abstract]” OR “Parenchymal Hemorrhage, Cerebral [Title/Abstract]” OR “Parenchymal Hemorrhages, Cerebral [Title/Abstract]” OR “Intracerebral Hemorrhage [Title/Abstract]” OR “Hemorrhage, Intracerebral [Title/Abstract]” OR “Hemorrhages, Intracerebral [Title/Abstract]” OR “Intracerebral Hemorrhages [Title/Abstract]” OR “Hemorrhage, Cerebral [Title/Abstract]” OR “Cerebral Hemorrhages [Title/Abstract]” OR “Hemorrhages, Cerebral [Title/Abstract]” OR “Brain Hemorrhage, Cerebral [Title/Abstract]” OR “Brain Hemorrhages, Cerebral [Title/Abstract]” OR “Cerebral Brain Hemorrhage [Title/Abstract]” OR “Cerebral Brain Hemorrhage Cerebral Brain Hemorrhages [Title/Abstract]” OR “Hemorrhage, Cerebral Brain [Title/Abstract]” OR “Hemorrhages, Cerebral Brain [Title/Abstract].”

#2 “ginkgo biloba” [MeSH Terms] OR “Ginkgo” [Title/Abstract] OR “Ginkgo biloba” [Title/Abstract] OR “Yinxing” [Title/Abstract] OR “Yinxingye” [Title/Abstract] OR “Shuxuening” [Title/Abstract] OR “Ginaton” [Title/Abstract].

#3 Clinical Trial [Publication Type] OR randomized controlled study [Title/Abstract] OR randomized controlled trial [Title/Abstract] OR randomized study [Title/Abstract] OR randomized trial [Title/Abstract] OR randomized placebo-controlled study [Title/Abstract] OR randomized placebo-controlled trial [Title/Abstract] OR randomized placebo controlled [Title/Abstract] OR randomized placebo-controlled [Title/Abstract] OR randomized double-blind [Title/Abstract] OR randomized double blin [Title/Abstract].

#4 #1 AND #2 AND #3.

Two reviewers independently conducted literature screening and data extraction and cross-checked in strict accordance with the eligibility criteria. If there was any disagreement, the third reviewer would discuss and solve the problem. The retrieved literature was imported into NoteExpress to exclude duplicate studies, then screening the titles and abstracts of the studies, and the full text was further screened. Data extraction was performed on the final included literature, including disease name, year, sample size, course of the disease, sex, intervention, course of treatment, outcome, and adverse events. Data extraction were independently completed and verified by two researchers, and disagreements were resolved through discussion or with the assistance of a third researcher.

The methodological quality of the included studies was assessed using the revised Cochrane Risk of Bias tool (ROB 2.0), including five areas: 1) randomization process (selection bias); 2) deviations from the intended interventions (performance bias); 3) missing outcome data (attrition bias); 4) measurement of the outcome (detection bias); 5) selection of the reported outcome (reporting bias) (Liu et al., 2021). The judgments for these domains were rated as three levels: low risk, some concerns and high risk. Finally, the overall bias was determined by summarizing the risk of bias across each domain. Risk of bias assessment were independently completed and verified by two researchers, and disagreements were resolved through discussion or with the assistance of a third researcher.

Statistical analysis was performed using Review Manager 5.3 and Stata 15.0. The dichotomous variables were expressed as risk ratios (RR), and the continuous variables were expressed as SMD or mean differences (MD) according to whether the measurement methods was different. The results were expressed as effect size with a 95% confidence interval (CI). Heterogeneity among studies was determined using the Chi² test and reported as I ² . When I ² was less than or equal to 50%, it was considered that the heterogeneity was low, and a fixed-effects model was used for pooling the results. Conversely, when I ² was greater than 50%, it was deemed that there was high heterogeneity, subgroup analysis and sensitivity analysis were used to identify the sources of heterogeneity. If the sources of heterogeneity could not be pinpointed, a random-effects model was employed for the meta-analysis. The efficacy and safety of interventional therapy might vary significantly depending on the timing of treatment initiation in patients with ICH, and 72 h represents a critical therapeutic time point (Zhou et al., 2012; Chinese Society of Neurology and Chinese Stroke Society, 2019; Li et al., 2024). Therefore, based on the disease course, patients were divided into two subgroups: the disease course ≥72 h group and the disease course <72 h group. Additionally, subgroup analyses would also be conducted based on differences in treatment duration and SXNI dosage across the groups. Funnel plots were used to assess the publication bias.

The certainty of evidence was assessed using the GRADE methodology, which categorizes it as high, moderate, low, or very low (Guyatt et al., 2008). Two reviewers independently evaluated the certainty of evidence for each outcome, with a third reviewer consulted to resolve any discrepancies. Detailed guidelines for assessing the certainty of evidence can be found in Supplementary Material S5.

Through database searches, a total of 1,946 studies were retrieved. Among them, 606 duplicate studies were excluded, and 1,340 studies were excluded because their titles and abstracts did not meet the eligibility criteria. The remaining 45 studies underwent full-text assessment. After full-text assessment, 16 studies were excluded due to the following reasons: 1) Unclear randomization (n = 5); 2) Not-RCT (n = 2); 3) Inconsistent interventions (n = 5); 4) No includable outcomes (n = 1); 5) The quality of the literature is extremely low (n = 3). Finally, 29 studies were included (Li et al., 2007; Liu, 2007; Wu, 2007; 2012; Liu and Liu, 2009; Zhang, 2009; 2011; Lan et al., 2010; Liu et al., 2010; Wang and Wang, 2010; Bai and Wang, 2012; Dong, 2012; Yuan et al., 2012; Zhang et al., 2012; 2015; Guo, 2013; Hu and Pu, 2015; Li and Hu, 2015; Wang, 2015; Wang et al., 2015; Min et al., 2016; Lu et al., 2018; Ni et al., 2018; Luo and Liu, 2019; Ma and Wang, 2019; Li, 2020; Teng, 2020; Zhong et al., 2021; Zuo et al., 2021) (Figure 1).

The total sample size of the 29 included studies was 3,012 participants, with 1,527 in the treatment group and 1,485 in the control group. These studies were published between 2007 and 2021, all in Chinese. These studies compared the efficacy and safety of SXNI combined with CT versus CT alone. The medications used in CT included oxiracetam injection, citicoline sodium, mannitol, edaravone injection, GM1, glycerol and fructose, furosemide, coenzyme A, inosine, β-aescin sodium. The average age of the patients ranged from 51.00 to 69.12, and the course of treatment ranged from 1 to 3 weeks. In the majority of the trials, males outnumbered females. No significant disparities were observed in the baseline characteristics across all the studies. The specific characteristics are shown in Table 1.

The 29 studies were assessed based on the five domains of ROB 2.0. In terms of the randomization process, 29 studies reported that the order of allocation was randomized, baseline balanced between groups, and were rated as some concern. Regarding deviations from intended interventions, the outcomes of 27 studies were not easily affected by the research environment and were rated as some concerns; 2 studies (Wang and Wang, 2010; Zhang et al., 2012) had outcomes that were more susceptible to bias due to environmental influences, which were rated as high risk. In terms of missing outcome data, none of the 29 studies reported any dropout cases, and all outcome data were complete, which was rated as low risk. For outcome measurement, although 27 studies did not describe blinding of assessors, their outcomes were not easily influenced by measurement, which was rated as low risk; 2 studies (Wang and Wang, 2010; Zhang et al., 2012) had outcomes that might be somewhat influenced by measurement, which was rated as some concerns. In terms of selective reporting of results, although none of the 29 studies had a predetermined plan, none of them reported selectively, which was rated as some concerns. Among the overall risk of bias, 27 studies were rated as some concerns, and 2 studies (Wang and Wang, 2010; Zhang et al., 2012) were rated as high risk. The results of the bias risk assessment for the included studies are shown in Figure 2.

By successively excluding literature, the combined results did not change significantly, indicating good stability of the outcomes. However, it was not possible to reduce the heterogeneity of the combined results by excluding literature. The graph of sensitivity analysis may be found in Supplementary Material S6.

Subgroup analysis was conducted based on different SXNI doses and durations of treatment. The low-dose subgroup was defined as those who received 10 mL or less of SXNI, while the high-dose subgroup received more than 10 mL. For the duration of treatment, the short-term treatment subgroup was defined as those treated for 14 days or less, and the long-term treatment subgroup was defined as those treated for more than 14 days. The results of each subgroup analysis were consistent with the overall results (Tables 3, 4). For patients with disease course ≤72h, the long-term treatment group was superior to the short-term treatment group in alleviating cerebral hematoma volume (p _interaction < 0.00001). In terms of reducing the cerebral hematoma volume, the low-dose subgroup outperformed the high-dose subgroup (p _interaction = 0.002). In terms of reducing cerebral edema volume, the high-dose subgroup was superior to the low-dose subgroup (p _interaction = 0.006). In improving ADL scores, the high-dose subgroup was superior to the low-dose subgroup (p _interaction = 0.001).

Funnel plots were used to assess publication bias for the four outcomes: neurological impairment score, cerebral hematoma volume, overall efficacy and cerebral edema volume. The funnel plots of neurological deficit score (Figure 14A), cerebral hematoma volume (Figure 14B), overall efficacy (Figure 14C) and cerebral edema volume (Figure 14D) showed obvious asymmetry, suggesting that there may be publication bias in the included studies.

Table 5 shows the assessment of the certainty of evidence. According to the results, the evidence for ESR, low cut whole blood viscosity, high cut whole blood viscosity, AE were considered to moderate certainty; HCT was considered to low certainty; neurological impairment score, overall efficacy, cerebral hematoma volume, cerebral edema volume, ADL score, and hs-CRP were considered as very low certainty.

Among the 10 outcomes included, some exhibited high heterogeneity. To address this, subgroup analyses were conducted on all outcomes based on treatment duration and the dosage of SXNI administered. And sensitivity analysis was also performed by sequentially excluding individual studies. Through sequential exclusion of individual studies, the meta-analysis results did not show significant alterations. Therefore, the sources of high heterogeneity in outcomes such as neurological impairment score, overall efficacy, cerebral edema volume, and hs-CRP remain unclear. It is speculated that the sources of heterogeneity may stem from factors such as the severity of intracerebral hemorrhage or patients’ comorbid conditions. This speculation requires further verification.

Due to methodological deficiencies in the included studies, there are some biases. Since none of the included studies reported allocation concealment, the risk of randomization process was judged as ‘some concerns’. Furthermore, selective reporting of results was rated as some concerns because none of the included studies reported trial registration. Overall, 93.1% of the studies were classified as some concerns regarding bias, with 6.9% classified as high risk. The assessment of publication bias indicated that four outcome measures may be subject to publication bias. This potential bias likely arises because certain types of studies (such as those with positive results, large effect sizes, or statistically significant findings) are more likely to be published, while studies with negative or null results may be overlooked or remain unpublished. Due to the methodological limitations of the included studies and potential publication bias, future high-quality studies are warranted to address these shortcomings and strengthen the current evidence base.

The GRADE approach was employed to evaluate the certainty of evidence in this study. Firstly, the risk of bias was conducted on the included studies. The study found that among the overall risks of bias for all outcomes, more than two-thirds were rated as issues of concern. Consequently, the certainty of all outcomes was downgraded by one level. Secondly, in terms of inconsistency in evidence, the heterogeneity test of 5 outcomes showed that I ² exceeding 75%, and the evidence was downgraded by two levels. And due to the heterogeneity test of one outcome showing I² > 50% and <75%, the evidence was downgraded by one level. Thirdly, there were no significant differences in the characteristics of the included studies. Therefore, the certainty of all outcomes was not downgraded due to indirect issues. Fourthly, in terms of imprecision, the certainty of an outcome is reduced by one level because the 95% confidence interval crossed the null line. The total sample size included in all studies was less than 400 cases, therefore, the certainty of one outcome was reduced by one level. Fifthly, regarding publication bias, the funnel plots of four outcomes showed significant asymmetry, suggesting potential publication bias. Consequently, the certainty of these four outcomes was reduced by one level. Due to downgrading, the certainty of the results in this study is affected, and therefore the results should be interpreted with caution.

The results of this study indicate that the combined use of SXNI and CT may significantly improve neurological function, promote the absorption of cerebral hematomas and cerebral edema, reduce inflammatory responses, improving microcirculatory blood flow, and demonstrated a favorable safety profile. These effects mightmay be related to the following pharmacological mechanisms of GBE: 1) Ginkgo leaf extract’s ability to regulate the inflammatory response of microglia, which in turn reduces the expression of inflammatory proteins in endothelial cells stimulated by oxidized low-density lipoprotein and decreases the release of inflammatory mediators by platelets, thereby mitigating the overall inflammatory response. And to alleviate brain damage by suppressing inflammatory responses (Sun et al., 2009; Kim et al., 2016; Wan et al., 2016; Chen et al., 2017). 2) GBE may also alleviate brain damage following cerebral hemorrhage by inhibiting neuronal apoptosis through the activation of the Akt signaling pathway (Ihl, 2013; Yu et al., 2018). 3) GBE protects human umbilical vein endothelial cells (HUVECs) from oxidized LDL-induced damage by inhibiting Akt phosphorylation, downregulating LOX-1 expression, and upregulating Sirt1 expression. Additionally, Ginkgo leaf extract may facilitate the relaxation of blood vessels, reduce blood pressure, and improve microcirculation by promoting the release of nitric oxide (NO) from endothelial cells (Kotıl et al., 2008; Wu et al., 2008; Ma et al., 2013; Silva and Martins, 2022). Under current therapeutic paradigms, Shuxuening demonstrates promising research and clinical application prospects as a potential multi-target agent for intracerebral hemorrhage management. Notably, the ≤72 h subgroup demonstrated superior therapeutic efficacy in cerebral edema reduction compared to the >72 h subgroup. Moreover, extended treatment duration or increased SXNI dosage may potentially enhance anti-inflammatory effects. Speculatively, this finding may be attributed to the following mechanisms: 1) The peak disruption of the blood-brain barrier (BBB) occurs within hours of ICH onset (Li Z. et al., 2020). Ginkgolides—the primary active components of SXNI—potently inhibit MMP-9 activation, a pivotal protease mediating early BBB damage (Fang et al., 2015). Early administration might therefore maximize cerebroprotective effects through BBB stabilization. 2) It might be associated with drug accumulation effects or exhibit a dose-dependent manner (Chen et al., 2022). Although the quality of evidence was low, this study still provides some references for the clinical use of SXNI in the treatment of cerebral hemorrhage, and provides ideas for further clinical research.

This study categorized patients into two phases: patients with disease course ≤72 h or >72h, and conducted a comprehensive evaluation of the efficacy and safety of SXNI across three aspects: clinical manifestations, imaging examinations, and laboratory tests. To explore the impact of various factors on treatment, the study also conducted subgroup analyses based on treatment duration and SXNI dosage. And discussions and speculations were made regarding the results of subgroup analyses.

This study has several limitations. First, the literature quality incorporated in this study was low, and the outcomes are susceptible to bias, diminishing the evidence’s credibility. Secondly, among the 29 literature included in this study, only literature from China and no literature from other countries were included, and all the included trials were conducted in China. This reduces the applicability of the evidence to other countries or regions. Third, the literature included in this study lacks follow-up and reports of end points, which may only represent the safety and efficacy of SXNI in the short term. Fourthly, due to the limited covariate information reported in the included literature, various subgroup analyses were conducted, yet the sources of significant heterogeneity in the results remain unclear.

The scarcity of well-conducted clinical trials is a prevalent concern within ethnopharmacology studies (Liu et al., 2024a). High-quality randomized controlled trials that involve substantial participant numbers and extended follow-up periods, characterized by meticulous design and execution, tend to demand greater financial and temporal investments, potentially accounting for the dearth of high-quality trial in included studies. To enhance the certainty of evidence regarding SXNI in cerebral hemorrhage, high-quality, large-sample, multicenter randomized controlled trials should be conducted to evaluate the efficacy and safety of SXNI in the treatment of cerebral hemorrhage. In the trial design, it is recommended to include outcomes that reflect long-term efficacy, such as mortality rates. It is also advised to use standardized assessment tools and methods, such as the NIHSS, to reflect specific neurological performance. And the pharmacological mechanisms of SXNI in the treatment of ICH still require further investigation.