Comparative efficacy of Honghua class injections for treating acute ischemic stroke: A Bayesian network meta-analysis of randomized controlled trials

Abstract

Background: Acute ischemic stroke (AIS) is associated with high morbidity, mortality, and disability. Clinical trials have shown that Honghua class injections (HCIs) combined with WM achieve better clinical efficacy than WM alone. In this study, we performed a Bayesian network meta-analysis (NMA) of randomized controlled trials (RCTs) to evaluate the efficacy of different HCIs combined with WM in treating AIS.

Methods: First, the inclusion and exclusion criteria were established. From inception to 1 June 2022, a systematic literature search was conducted in multiple databases for the treatment of AIS with HCIs, including Honghua injection (HI), Safflower Yellow injection (SYI), Guhong injection (GHI), and Danhong injection (DHI). Subsequently, OpenBUGS 3.2.3 was applied to conduct a Bayesian algorithm, and Stata 16.0 was used to prepare the graphs. Multidimensional cluster analysis was performed using the “scatterplot3d” package in R 3.6.1 software.

Results: In this NMA, a total of 120 eligible RCTs were included, involving 12,658 patients, and evaluating the clinical effectiveness rates, activities of daily living (ADL), hemorheological indexes, and adverse reactions (ADRs). DHI + WM was the best intervention for improving the clinical effectiveness rate. Moreover, cluster analysis demonstrated that DHI + WM and SYI + WM had better comprehensive therapeutic effects. As most of the included RCTs did not monitor ADRs, the safety of the HCIs remains to be further explored.

Conclusion: DHI + WM and SYI + WM probably have a better clinical efficacy on AIS patients. Nevertheless, due to the limitation of this NMA, this conclusion may be biased. High-quality RCTs should be performed to validate our findings.

Systematic Review Registration:https://www.crd.york.ac.uk/PROSPERO/, identifier CRD42021229599

1 Introduction

Stroke is the leading cause of death in China and the second-most prevalent cause of death worldwide (Benjamin et al., 2018; Zhou et al., 2019). The prevalence of stroke in China has been increasing continuously since 2006, with ∼13 million stroke patients. Ischemic stroke (IS) has the highest incidence, accounting for ∼80% of all stroke patients (Benjamin et al., 2018). Due to its high morbidity, mortality, and disability, stroke has become a major disease that seriously endangers human health. At present, the treatment of acute ischemic stroke (AIS) mainly includes thrombolysis, intervention, anti-platelet aggregation, anticoagulation, lowering the levels of fibrinogen and blood lipids, expansion of blood capacity, and neuroprotection. Notably, the most effective treatment is ultra-early thrombolysis (Ospel et al., 2020). However, due to the strict treatment time window of thrombolysis, the population in which this treatment is carried out is extremely limited in China (Zhu et al., 2020). Therefore, it is necessary to explore other effective therapeutic methods. Traditional Chinese medicine (TCM) has several advantages, such as multiple targets, good synergy, and low side effects, widely used in treating complex diseases (Chen et al., 2017). TCM-related injections (TCMIs) are often combined with western medicine (WM) for the treatment of acute diseases in Chinese clinics and show instant effectiveness and high bioavailability (Li et al., 2017). The detailed pharmacology information, composition and the extraction procedure of HCIs have been stated unambiguously, shown in Supplementary Material S1.

In TCM theory, AIS is usually related to blood stasis syndrome and its treatment strategy involves the promotion of blood circulation and removal of blood stasis (Tang et al., 2012; Wang et al., 2020b). Honghua (Asteraceae, Carthamus, Carthamus tinctorius L.) is commonly used in the treatment of ischemic cardiovascular and cerebrovascular diseases, as it has a good effect on promoting blood circulation and removing blood stasis, which was described in the Compendium of Materia Medica (Ming Dynasty, ∼500 years ago) (Cao et al., 2014; Bai et al., 2020). Moreover, Hydroxysafflor yellow A is the main active component of Honghua (Bai et al., 2020). TCMIs contain an extract from Honghua, including Honghua injection (HI), Safflower Yellow injection (SYI), Guhong injection (GHI), and Danhong injection (DHI). Herein, we call them Honghua class injections (HCIs). All HCIs have been approved by the State Pharmaceutical Administration of China. HCIs combined with WM are commonly adopted in treating AIS and achieve a good clinical effect (Li et al., 2015b; Du et al., 2018; Feng et al., 2018).

Compared with traditional meta-analyses, network meta-analysis (NMA) can synthesize multiple interventions and perform direct or indirect comparisons for the same disease (Guo et al., 2020). Moreover, it can help to evaluate and rank the efficacy of different treatments (Catalá-López et al., 2014). Existing studies demonstrated that HCIs combined with WM achieved better clinical efficacy than WM alone in the treatment of AIS. However, there is a lack of clinical trials comparing HCIs directly. Therefore, it is necessary to evaluate and compare the efficacy of various HCIs by NMA. In this study, four HCIs, including HI, SYI, GHI, and DHI were selected as adjuvant therapies for AIS, which were all combined with conventional WM treatments. Subsequently, we applied a Bayesian NMA to explore the comparative effectiveness and safety between different HCIs combined with WM against AIS, providing a reference for clinical practice.

2 Methods

This NMA has been registered on the International Prospective Register of Systematic Review (PROSPERO) platform (CRD42021229599). This NMA study was conducted strictly according to the guidelines based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) as shown in Supplementary Material S2 (Moher et al., 2009).

2.1 Search strategy

All RCTs focusing on HCIs against AIS literatures were searched electronically from the following seven databases, Cochrane Library, PubMed, China National Knowledge Infrastructure Database (CNKI), China Biomedical Literature Service System (SinoMed), and Wan-Fang Database. All database searches were conducted on studies dating from inception to 1 June 2022, with no restrictions on language. The Medical Subject Heading (MeSH) terms and free-text keywords were utilized, including “acute ischemic stroke (MeSH Terms),” “ischemic stroke,” “acute ischemic stroke,” “stoke,” “brain infarction,” “acute cerebral infarction,” “cerebral infarction,” “brain embolism,” “cerebrovascular disorders,” “Honghua injection,” “Safflower Yellow injection,” “Guhong injection,” “Danhong injection,” and “randomized controlled trial (Publication Type).” In addition, there were no restrictions on the blinding methods, publication year, and language. Furthermore, the specific retrieval strategy is shown in Supplementary Material S3.

2.2 Inclusion criteria

2.2.1 Patient populations

All included cases were diagnosed with AIS. This study only recruited patients within 2 weeks of onset. There were no restrictions on the age, gender, race, and severity of disease.

2.2.2 Interventions and comparators

The interventions of experiment groups were HCIs (HI, SYI, GHI, or DHI) combined with WM treatments. The control group only received WM therapy. Conventional WM treatment, including anti-platelet aggregation; anticoagulation; lipid-lowering; correction of water, electrolyte disorders, and acid-base imbalance; improvement of cerebral circulation and using neuroprotective agent. The dosage and duration of treatment was not restricted.

2.2.3 Outcomes

The primary outcome was the clinical effectiveness rate, according to the National Institute of Health Stroke Scale (NIHSS) evaluation: a reduction of 91%–100%, 46%–90%, and 18%–45% corresponds to “basic cure,” “notable progress,” and “progress” respectively, which defined the effectiveness (Wang et al., 2018).

The secondary outcomes were the activities of daily living (ADLs), using Barthel Index Scale Scores; hemorheological indexes, including low shear blood viscosity (LBV), high shear blood viscosity (HBV), plasma viscosity (PV), fibrinogen (FIB) levels; and adverse reactions (ADRs), including adverse drug events (ADEs).

2.2.4 Study design

Randomized controlled trials (RCTs) of HCIs combined with WM in the treatment of AIS. RCTs were not restricted by language, country, publication date, or stage.

2.3 Exclusion criteria

The RCTs that met one of the following conditions were excluded: 1) RCTs that did not meet the criteria for clinical efficacy evaluation; 2) thrombolytic therapy; 3) interventions involving a combination therapy with Chinese herbal medicine, acupuncture or other TCM injections; 4) the full text of the study was unavailable; 5) incorrect or incomplete data; 6) duplicate reports; and 7) no related outcomes.

2.4 Data extraction

All the articles were managed by NoteExpress software (Tongji University Library, Shanghai, China) and selected by two independent reviewers by excluding irrelevant articles, reviews, and animal experiments. Two reviewers independently extracted the eligible research data using Excel (Microsoft, United States) and the collected the information as follows: 1) sample size in each group, sex, age, and duration of disease; 2) intervention (the types of HCIs, dose, and course of treatment); 3) outcome indicators: clinical effectiveness rate, ADL, LBV, HBV, PV, FIB levels, and ADRs/ADEs; 4) factors to evaluate the risk of bias.

2.5 Quality assessment

In this study, we applied the Cochrane bias risk assessment tool to assess the methodological quality of the included RCTs (Higgins et al., 2011). The tool assessment items were as follows: random sequence generation, allocation concealment, blinding of participants and researchers, blinding of outcome assessment, incomplete outcome data, selective reporting, and other bias. There were two independent investigators (LL and CS) who assessed the quality of the included RCTs. Any disagreement was resolved by a third researcher (HW) or by consensus. RevMan 5.4 software was applied to generate the risk of bias diagram for quality evaluation.

2.6 Statistical analysis

Bayesian modeling was performed using OpenBUGS 3.2.3, (MRC Biostatistics Unit, Cambridge, UK) (Owen et al., 2020). Each chain in the OpenBUGS program has 50,000 iterations, with the first 20,000 iterations being burn-in tests to remove the initial value effect. The Odds ratio (OR) and mean difference (MD) with a 95% credible interval (CI) were used to estimate the binary results and continuous data, respectively (Da Costa et al., 2013; Pateras et al., 2018). If the ORs did not include 1 and MDs did not cover 0, the difference between the two groups was deemed significant (Da Costa et al., 2013; Pateras et al., 2018).

The software, Stata 16.0 (College Station, TX, United States) was applied to generate graphs, carry out the publication bias test, and consistency test. In the network graphs, nodes represented interventions, and the link lines indicated direct comparisons. For each outcome, the probability of the intervention was ranked using the surface under the cumulative ranking area (SUCRA), a larger area under the curve indicated a better cure (Zhou et al., 2020). Based on SUCRA value, R 3.6.1 software (Mathsoft, Cambridge, United States) was applied to perform the cluster analysis using K-means method (Guo et al., 2020). In addition, “scatterplot3d” package was used for multidimensional cluster analysis. The comprehensive curative effects of HCIs in two or three outcomes were evaluated.

The publication bias was assessed using a comparison-adjusted funnel plot (Song et al., 2010). Publication bias did not exist if the funnel plots were symmetrical (Debray et al., 2018). Moreover, as this NMA had no closed loops, the overall consistency test could not be performed. However, a local consistency test was performed. It was determined that there were no local inconsistencies in this study when p > 0.05.

3 Results

3.1 Literature selection

Using the search strategy, a total of 2,669 articles were retrieved. After removing duplicates, 2,619 articles were obtained. Subsequently, 2,481 articles were excluded on account of being reviews, meta-analyses, systematic reviews, animal experiments, and other irrelevant literature, or on account of having inconsistencies in inclusion criteria. A total of 158 articles were evaluated for qualification. Next, 38 articles were excluded for the following reasons: thrombolytic therapy, full text of the study was unavailable, incorrect or incomplete data, duplicate reports, and no related outcomes. Finally, 120 RCTs were eligible for inclusion in this NMA (Figure 1).

FIGURE 1

3.2 Study characteristics

Four types of HCIs were incorporated, including HI, SYI, GHI, and DHI. Overall, the 120 RCTs involved 12,658 patients (6,185 in the control group and 6,473 in the experiment group). A total of four comparisons were evaluated: HI + WM vs. WM (n = 15), SYI + WM vs. WM (n = 8), GHI + WM vs. WM (n = 12), and DHI + WM vs. WM (n = 85). The control groups were treated with WM, which primarily contained aspirin, anticoagulants, neuroprotectants, etc. The characteristic details of the RCTs were shown in Table 1, and the included literatures are described in Supplementary Material S4. The network graphs of the four HCIs with different outcomes were shown in Figure 2.

FIGURE 2

3.3 Quality evaluation

The risk of bias in RCTs included in this NMA was assessed using a tool developed by the Cochrane Collaboration (Higgins et al., 2011). In 35 RCTs, random sequences were generated using a random number table, considered as low risk in terms of selection bias; the allocation concealment was not clear. In terms of performance bias, four RCTs mentioned blinding, evaluated as low risk, and one RCT clearly pointed out that blinding was not used, considered to be high risk. In all RCTs, complete data were available, and the attrition bias was evaluated as low risk. Moreover, reporting bias and other bias were not clear. The risk of bias for the RCTs included in this NMA is described in Figure 3.

FIGURE 3

3.4 Outcomes

3.4.1 Clinical effectiveness rates

A total of 104 RCTs compared the clinical effectiveness rates: HI + WM vs. WM (n = 14), SYI + WM vs. WM (n = 7), GHI + WM vs. WM (n = 8), and DHI + WM vs. WM (n = 75). As shown in Figure 4A, HCIs + WM treatments had significantly higher clinical effectiveness rates than WM alone: HI + WM (OR = 0.26, 95% CI: 0.19, 0.36), SYI + WM (OR = 0.26, 95% CI: 0.15, 0.40), GHI + WM (OR = 0.3, 95% CI: 0.18, 0.46), and DHI + WM (OR = 0.25, 95% CI: 0.22, 0.29). In addition, the results of a pairwise comparison of the four injections were as follows: HI + WM vs. SYI + WM (OR = 1.09, 95% CI: 0.60,1.81), HI + WM vs. GHI + WM (OR = 0.93, 95% CI: 0.52,1.56), HI + WM vs. DHI + WM (OR = 1.06, 95% CI: 0.73, 1.47), SYI + WM vs. GHI + WM (OR = 0.91, 95% CI: 0.43, 1.64), and GHI + WM vs. DHI + WM (OR = 1.2, 95% CI: 0.73, 1.88), indicating there were no significantly differences in the pairwise comparisons of the four types of HCI.

FIGURE 4

The SUCRA is shown in Figures 4A,B, DHI + WM (71.08%) >SYI + WM (67.89%) >HI + WM (62.88%) >GHI + WM (48.16%) >WM (0%). According to the SUCRA, DHI + WM treatment was the best intervention for improving the clinical effectiveness rate.

3.4.2 ADL Function

A total of 36 RCTs investigated ADL: HI + WM vs. WM (n = 2), SYI + WM vs. WM (n = 5), GHI + WM vs. WM (n = 4), and DHI + WM vs. WM (n = 23). Figure 5B shows the comparisons of SYI + WM vs. WM (MD = −17.68, 95% CI: −25.52,−9.83) and DHI + WM vs. WM (MD = −15.36, 95% CI: 19.26, −11.45). There were no significantly differences in the pairwise comparisons of other types of HCI.

FIGURE 5

The SUCRA is shown in Figures 5A,B, SYI + WM (78.81%), DHI + WM (65.31%), HI + WM (60.93%), and GHI + WM (41.91%), respectively. Therefore, SYI + WM treatment was probably the best effective intervention for improving theADL.

3.4.3 LBV level

A total of 27 RCTs investigated the LBV: HI + WM vs. WM (n = 1), SYI + WM vs. WM (n = 3), GHI + WM vs. WM (n = 8), and DHI + WM vs. WM (n = 15). Figure 6C shows the comparisons of GHI + WM vs. WM (MD = 1.14, 95% CI: 0.59, 1.67) and DHI + WM vs. WM (MD = 0.56, 95% CI: 0.15, 0.98).

FIGURE 6

The SUCRA is shown in Figures 6A,C, GHI + WM (76.54%) > HI + WM (63.83%) > SYI + WM (62.43%) > DHI + WM (41.36%). According to the ranking of SUCRA, the GHI + WM (76.54%) treatment was probably to be the best intervention in neurological impairment.

3.4.4 HBV level

A total of 26 RCTs investigated the HBV level: HI + WM vs. WM (n = 1), SYI + WM vs. WM (n = 3), GHI + WM vs. WM (n = 6), and DHI + WM vs. WM (n = 16). All CHIs combined with WM achieved a better effect in HBV than using WM alone. The significant results have been shown in Figure 6D, HI + WM vs. WM (MD = 1.62, 95% CI: 0.07, 3.07), SYI + WM vs. WM (MD = 1.43, 95% CI: 0.61, 2.39), GHI + WM vs. WM (MD = 1.41, 95% CI: 0.22, 2.58), and DHI + WM vs. WM (MD = 0.98, 95% CI: 0.38, 1.63).

The SUCRA is shown in Figures 6B,D, HI + WM (74.01%), SYI + WM (68.33%), GHI + WM (65.64%), and DHI + WM (41.15%), respectively. According to the SUCRA probabilities, the HI + WM treatment appeared to be the best interventions for HBV.

3.4.5 PV level

A total of 33 RCTs investigated the PV: HI + WM vs. WM (n = 1), SYI + WM vs. WM (n = 3), GHI + WM vs. WM (n = 4), and DHI + WM vs. WM (n = 25). The statistically significant results are shown in Figure 7C, DHI + WM vs. WM (MD = 0.38, 95% CI: 0.20, 0.57).

FIGURE 7

The SUCRA is shown in Figures 7A,C, GHI + WM (73.91%), SYI + WM (69.77%), DHI + WM (57.42%), and HI + WM (39.27%). GHI + WM treatment was probably to be the best intervention in plasma viscosity. However, GHI + WM vs. WM (MD = 0.55, 95% CI: −0.06, 1.17) and SYI + WM vs. WM (MD = 0.50, 95% CI: −0.06, 1.06), indicating there was no significant difference in the pairwise comparisons.

3.4.6 FIB level

A total of 33 RCTs investigated the FIB level: HI + WM vs. WM (n = 0), SYI + WM vs. WM (n = 2), GHI + WM vs. WM (n = 4), and DHI + WM vs. WM (n = 27). The statistically significant results are shown in Figure 7D, DHI + WM vs. WM (MD = 0.87, 95% CIs: 0.57, 1.2), and GHI + WM vs. WM (MD = 0.55, 95% CIs: 0.02, 1.10).

The SUCRA is shown in Figures 7B,D, DHI + WM (88.90%), GHI + WM (56.09%), and SYI + WM (51.41%). Accordingly, DHI + WM treatment was the best intervention in decreasing the FIB level.

3.5 Cluster analysis

When cluster analysis was performed to four HCIs that reported the clinical effectiveness rate and ADL, we found SYI + WM, DHI + WM and HI + WM were classified into the same category, indicating they exerted similar effectiveness (Figure 8A). In addition, we performed multidimensional cluster analysis on interventions with more than 1 included RCTs to further evaluate the comprehensive efficacy of HCIs. All the results demonstrated DHI + WM and SYI + WM might have better therapeutic effects (Figures 8B–E).

FIGURE 8

3.6 Safety evaluation

Of the 120 RCTs, 59 trials investigated the ADRs/ADEs, involving a total of 5,749 patients. Overall, 2,778 cases in the control group (WM), and 2,971 cases in the experiment group (HCIs + WM) were included.

In terms of the safety evaluation of HI, six RCTs were included, involving 674 patients, 320 cases in the WM group and 354 cases in the HI + WM group. There were three cases of ADRs that occurred in the HI + WM group: pruritus (2 cases) and skin flushing (1 case).

In terms of the safety evaluation of SYI, five RCTs were included, involving 554 patients, 277 cases in the WM group and 277 cases in the SYI + WM group. There were seven cases of ADRs that occurred in the SYI + WM group: rash (2 cases), gastrointestinal reactions (4 cases), such as nausea, vomiting and diarrhea, and gingival bleeding (1 case). Five cases of ADRs occurred in the WM group: gastrointestinal reactions (3 cases) and gingival bleeding (2 cases).

In terms of the safety evaluation of GHI, seven RCTs were included, involving 1055 patients, 514 cases in the WM group, and 541 cases in the GHI + WM group. There were 23 cases of ADRs in the GHI + WM group: 10 cases of gastrointestinal reactions, such as nausea, vomiting, and diarrhea, cases of headache or dizziness (5 cases), rash (3 cases), vasculitis (1 case), lethargy (2 cases), and mild fever (2 cases). 29 cases of ADRs occurred in the WM group: gastrointestinal reactions (11 cases), headache or dizziness (8 cases), rash (1 case), gingival bleeding (2 cases), vasculitis (2 cases), 5 cases of lethargy, and mild abnormal liver function (2 cases).

In terms of the safety evaluation of DHI, 41 RCTs were included, involving 3,847 patients, 1,848 cases in the WM group, and 1,999 cases in the DHI + WM group. There were 53 cases of ADRs that had occurred in the DHI + WM group: gastrointestinal reactions (18 cases), headache or dizziness (19 cases), rash (7 cases), vasculitis (1 case), fatigue (1 case), 1 case of hypotension, mild palpitation (2 cases), limb pain (2 cases), and cough (2 cases). 64 cases of ADRs occurred in the WM group: gastrointestinal reactions (17 cases), headache or dizziness (19 cases), rash (9 cases), vasculitis (2 cases), fatigue (1 case), hypotension (3 cases), mild palpitation (3 cases), epigastric discomfort (4 cases), limb pain (3 cases), and cough (3 cases).

3.7 Publication bias

The funnel plot of the clinical effectiveness rate was not quite symmetrical, indicating the potential publication bias in this NMA (Figure 9).

FIGURE 9

3.8 Consistency test

In this NMA, as there were no closed loops, an overall consistency test was not possible. All the evidences about these contrasts were obtained from the trials, which directly compare them. The local consistency test showed that p > 0.05, indicating that there was no local inconsistency in this study.

4 Discussion

We performed a NMA on four common HCIs and compared the outcomes to determine the most appropriate choice, providing some references for the clinical treatment of AIS. This NMA included 120 RCTs involving 12,658 patients, evaluating the clinical effectiveness rate, ADL, LBV HBV, PV, FIB levels, and ADRs after the application of four HCIs combined with WM and WM alone. The clinical effectiveness rate, evaluated by neurological function recovery, was considered as the primary outcome. According to the ranking of SUCRA, DHI + WM treatment was the best intervention for improving the clinical effectiveness rate, while SYI + WM was the best treatment for ADL. Moreover, multidimensional cluster analysis demonstrated that DHI + WM and SYI + WM might have a better comprehensive therapeutic effect.

DHI is extracted from Danshen (Salvia miltiorrhiza Bunge, Lamiaceae, Salviae miltiorrhizae radix et rhizoma) and Honghua, widely used in the treatment of AIS. The main active components of DHI are danshensu, protocatechuic aldehyde, safflower yellow A, and salvianolic acid (Jiang et al., 2015; Liu et al., 2019). The clinical trial has shown that the value of DHI in treating AIS with blood stasis syndrome is comprehensively evaluated as Grade A (Cui et al., 2021). Moreover, pharmacological studies have shown that DHI can protect the blood-brain barrier after AIS (Zeng et al., 2021a), improve the energy metabolism of cells in the ischemic area (Zeng et al., 2021b), enhance the mitochondrial function (Orgah et al., 2019), exert anti-inflammatory, anti-apoptotic, and antioxidant effects (Wang et al., 2020a), thereby improving neurological impairment. SYI is extracted from Honghua, its main component is safflower yellow. A clinical trial on IS has shown that SYI may treat AIS by alleviating the inflammation reaction of patients (Li et al., 2015a). The main active component of safflower yellow is hydroxysafflower yellow A (HYSA), with antioxidant activity. Pharmacological studies have demonstrated that HYSA plays an important role in anti-atherosclerosis (Xue et al., 2021), protecting neurons from excitotoxic damage (Wang et al., 2016), dilating cerebral vessels, and improving the cerebrovascular permeability (Sun et al., 2018; Li et al., 2022).

Atherosclerosis is the pathological basis of IS. Abnormal hemorheological indexes, especially the increase of plasma viscosity and fibrinogen levels, will accelerate the pathological process of atherosclerosis and lead to stroke. Furthermore, a previous study has shown that hemorheological disturbances affect cognitive functions of IS patients (Velcheva and Nikolova, 2008). Ameliorating hemorheology is of great significance to the prognosis of stroke (Resch, 1992). In this NMA, we found that GHI had the best effect in reducing PV and LBV and DHI had the best effect in reducing FIB. As there was only one RCT in HI treatment, it could not prove that the curative effect of HI was better than other HCIs. GHI comprised safflower aqueous extract and aceglutamide. Several studies have shown that the mechanisms of GHI in treating IS were related to inhibiting inflammation and reducing neuronal apoptosis (Wang et al., 2022; Zhang et al., 2022). During inflammation, the hemorheological system is impaired (Pretorius, 2018). The regulation of hemodynamics by GHI may be related to its anti-inflammatory effects.

In addition to the clinical efficacy, the safety of CHIs in treating AIS is also particularly important. In this NMA, 59 RCTs investigated the ADRs/ADEs (HI 6 RCTs; SYI five RCTs; GHI seven RCTs and DHI 41 RCTs), more than 50% of RCTs did not report safety. However, more than 50% of RCTs did not report them. Therefore, we were unable to make the accurate conclusions of HCIs’ safety. Previous studies had shown that most ADRs of DHI were mild and moderate. Post-marketing safety monitoring and re-evaluation studies of DHI with 30,888 cases show that the incidence of ADRs/ADEs was 3.50‰ (Li et al., 2015b). Unfortunately, there is a lack of large samples and high-quality safety research on other HCIs post-marketing. Since there are few studies focused on safety assessments in this NMA, further experimental and clinical evidence is required to verify the safety of these HCIs. Moreover, in order to improve the safety of clinical applications of TCMIs, a research system and post-marketing safety surveillance and re-evaluation should be established.

This study is a NMA based on RCTs, using a Bayesian algorithm, which has some advantages. A comprehensive multi-platform literature search was conducted for this study and strict inclusion and exclusion criteria were formulated. Importantly, this is the first NMA to evaluate the clinical efficacy of HCIs + WM in treating AIS. We also analyzed the changes of hemorheological indexes and the safety of HCIs, which has certain significance for guiding the treatment of AIS.

However, this study still has some limitations. First, the current NMA-included RCTs were observed to have a selection bias and performance bias; there was a lack of information regarding allocation concealment and blinding of participants or personnel in most RCTs. Second, apart from DHI + WM, there were few eligible RCTs for other interventions, which might lead to bias of the outcomes. Third, most of the included RCTs were single-center studies, and the patients included in the studies may be regionalized. Additionally, all of the RCTs in this NMA were conducted among Chinese population. Whether the curative effect is affected by race or region is still unclear. Importantly, clinical heterogeneity might happen due to the diversity of conventional WM treatment and the different time of onset, dosage and the course of treatment. Moreover, few eligible RCTs reported follow-up results; the recurrence and mortality rates of patients treated with HCI + WM are not well understood. In view of the above limitations, RCTs should be conducted more standardized, clearly defined in terms of populations, interventions, comparators, outcomes, and study designs (PICOS). In order to accurately evaluate the efficacy and safety of TCMIs and promote their application in the real-world clinical practice, the methodological quality of RCTs should be improved.

5 Conclusion

In conclusion, based on the results of this Bayesian NMA, HCIs combined with WM treatments significantly improve the therapeutic effect, and DHI and SYI combined with WM treatments are probably preferred among HCIs for the treatment of AIS. However, due to the limitations, this conclusion may be biased. Importantly, high-quality, multicenter, and double-blind RCTs should be performed in the future to validate our findings. Additionally, we need to improve the quality of TCMI security assessment in RCTs, strictly monitor the ADRs/ADEs of TCMI, and standardize medication.

References

• 1

  BaiX.WangW.FuR.YueS.GaoH.ChenY.et al (2020). Therapeutic potential of hydroxysafflor yellow A on cardio-cerebrovascular diseases. Front. Pharmacol.11, 01265. 10.3389/fphar.2020.01265

  • CrossRef
  • Google Scholar

• 2

  BenjaminE. J.ViraniS. S.CallawayC. W.ChamberlainA. M.ChangA. R.ChengS.et al (2018). Heart disease and stroke statistics-2018 update: A report from the American heart association. Circulation137, e67–e492. 10.1161/CIR.0000000000000558

  • CrossRef
  • Google Scholar

• 3

  CaoJ.ChenZ.ZhuY.LiY.GuoC.GaoK.et al (2014). Huangqi-Honghua combination and its main components ameliorate cerebral infarction with Qi deficiency and blood stasis syndrome by antioxidant action in rats. J. Ethnopharmacol.155, 1053–1060. 10.1016/j.jep.2014.05.061

  • CrossRef
  • Google Scholar

• 4

  Catalá-LópezF.TobíasA.CameronC.MoherD.HuttonB. (2014). Network meta-analysis for comparing treatment effects of multiple interventions: An introduction. Rheumatol. Int.34, 1489–1496. 10.1007/s00296-014-2994-2

  • CrossRef
  • Google Scholar

• 5

  ChenH. S.QiS. H.ShenJ. G. (2017). One-Compound-Multi-Target: Combination prospect of natural compounds with thrombolytic therapy in acute ischemic stroke. Curr. Neuropharmacol.1, 134–156. 10.2174/1570159x14666160620102055

  • CrossRef
  • Google Scholar

• 6

  CuiX.HanS.JinX. L.WangZ. F.ZhangQ.XieY. M. (2021). Clinical comprehensive evaluation of Danhong Injection in treatment of stroke with blood stasis syndrome. Zhongguo Zhong Yao Za Zhi46, 6096–6104. 10.19540/j.cnki.cjcmm.20210930.506

  • CrossRef
  • Google Scholar

• 7

  Da CostaB. R.NüeschE.RutjesA. W.JohnstonB. C.ReichenbachS.TrelleS.et al (2013). Combining follow-up and change data is valid in meta-analyses of continuous outcomes: A meta-epidemiological study. J. Clin. Epidemiol.66, 847–855. 10.1016/j.jclinepi.2013.03.009

  • CrossRef
  • Google Scholar

• 8

  DebrayT. P. A.MoonsK. G. M.RileyR. D. (2018). Detecting small-study effects and funnel plot asymmetry in meta-analysis of survival data: A comparison of new and existing tests. Res. Synth. Methods9, 41–50. 10.1002/jrsm.1266

  • CrossRef
  • Google Scholar

• 9

  DuY.ZhongL.WuS.WangX. (2018). Comparative study of the effects of Danhong injection with different doses on ischemic stroke: A substudy of hospital-based Danhong injection registry. J. Traditional Chin. Med.38, 917–925. 10.1016/s0254-6272(18)30992-0

  • CrossRef
  • Google Scholar

• 10

  FengW. W.ZhangY.TangJ. F.ZhangC. E.DongQ.LiR. Y.et al (2018). Combination of chemical fingerprinting with bioassay, a preferable approach for quality control of Safflower Injection. Anal. Chim. Acta1003, 56–63. 10.1016/j.aca.2017.11.069

  • CrossRef
  • Google Scholar

• 11

  GuoS.WuJ.NiM.JiaS.ZhangJ.ZhouW.et al (2020). Comparative efficacy of danshen class injections for treating acute coronary syndrome: A multidimensional bayesian network meta-analysis of randomized controlled trials. Front. Pharmacol.11, 1260. 10.3389/fphar.2020.01260

  • CrossRef
  • Google Scholar

• 12

  HigginsJ. P.AltmanD. G.GøtzscheP. C.JüniP.MoherD.OxmanA. D.et al (2011). The Cochrane Collaboration's tool for assessing risk of bias in randomised trials. Bmj343, d5928. 10.1136/bmj.d5928

  • CrossRef
  • Google Scholar

• 13

  JiangX.LvB.LiP.MaX.WangT.ZhouQ.et al (2015). Bioactivity-integrated UPLC/Q-TOF-MS of Danhong injection to identify NF-κB inhibitors and anti-inflammatory targets based on endothelial cell culture and network pharmacology. J. Ethnopharmacol.174, 270–276. 10.1016/j.jep.2015.08.026

  • CrossRef
  • Google Scholar

• 14

  LiJ. P.LiuY.GuoJ. M.ShangE. X.ZhuZ. H.ZhuK. Y.et al (2017). A comprehensive strategy to evaluate compatible stability of Chinese medicine injection and infusion solutions based on chemical analysis and bioactivity assay. Front. Pharmacol.8, 833. 10.3389/fphar.2017.00833

  • CrossRef
  • Google Scholar

• 15

  LiL. J.LiY. M.QiaoB. Y.JiangS.LiX.DuH. M.et al (2015a). The value of safflower yellow injection for the treatment of acute cerebral infarction: A randomized controlled trial. Evid. Based. Complement. Altern. Med.2015, 478793. 10.1155/2015/478793

  • CrossRef
  • Google Scholar

• 16

  LiX. L.TangJ. F.LiW. X.LiC. X.ZhaoT.ZhaoB. C.et al (2015b). Postmarketing safety surveillance and reevaluation of Danhong injection: Clinical study of 30888 cases. Evid. Based. Complement. Altern. Med.2015, 610846. 10.1155/2015/610846

  • CrossRef
  • Google Scholar

• 17

  LiY.LiuX. T.ZhangP. L.LiY. C.SunM. R.WangY. T.et al (2022). Hydroxysafflor yellow A blocks HIF-1α induction of NOX2 and protects ZO-1 protein in cerebral microvascular endothelium. Antioxidants (Basel)11, 728. 10.3390/antiox11040728

  • CrossRef
  • Google Scholar

• 18

  LiuS.WangK.DuanX.WuJ.ZhangD.LiuX.et al (2019). Efficacy of danshen class injection in the treatment of acute cerebral infarction: A bayesian network meta-analysis of randomized controlled trials. Evid. Based. Complement. Altern. Med.2019, 5814749. 10.1155/2019/5814749

  • CrossRef
  • Google Scholar

• 19

  MoherD.LiberatiA.TetzlaffJ.AltmanD. G. (2009). Preferred reporting items for systematic reviews and meta-analyses: The PRISMA statement. Ann. Intern. Med.151, 264–269. 10.7326/0003-4819-151-4-200908180-00135

  • CrossRef
  • Google Scholar

• 20

  OrgahJ.RenJ.LiuX.OrgahE.GaoX.ZhuY. J. R. N.et al (2019). Danhong injection facilitates recovery of post-stroke motion deficit via Parkin-enhanced mitochondrial function. Restor. Neurol. Neurosci.37, 375–395. 10.3233/RNN-180828

  • CrossRef
  • Google Scholar

• 21

  OspelJ. M.MenonB. K.DemchukA. M.AlmekhlafiM. A.KashaniN.MayankA.et al (2020). Clinical course of acute ischemic stroke due to medium vessel occlusion with and without intravenous alteplase treatment. Stroke51, 3232–3240. 10.1161/STROKEAHA.120.030227

  • CrossRef
  • Google Scholar

• 22

  OwenR. K.BujkiewiczS.TincelloD. G.AbramsK. R. (2020). Multivariate network meta-analysis incorporating class effects. BMC Med. Res. Methodol.20, 184. 10.1186/s12874-020-01025-8

  • CrossRef
  • Google Scholar

• 23

  PaterasK.NikolakopoulosS.RoesK. (2018). Data-generating models of dichotomous outcomes: Heterogeneity in simulation studies for a random-effects meta-analysis. Stat. Med.37, 1115–1124. 10.1002/sim.7569

  • CrossRef
  • Google Scholar

• 24

  PretoriusE. (2018). Erythrocyte deformability and eryptosis during inflammation, and impaired blood rheology. Clin. Hemorheol. Microcirc.69, 545–550. 10.3233/CH-189205

  • CrossRef
  • Google Scholar

• 25

  ReschK. L. (1992). Hemorheology and stroke. Fortschr. Med.110, 101–104.

  • Google Scholar

• 26

  SongF.ParekhS.HooperL.LokeY. K.RyderJ.SuttonA. J.et al (2010). Dissemination and publication of research findings: An updated review of related biases. Health Technol. Assess.14, 1–193. iii, ix-xi. 10.3310/hta14080

  • CrossRef
  • Google Scholar

• 27

  SunY.XuD.QinZ.WangP.HuB.YuJ.et al (2018). Protective cerebrovascular effects of hydroxysafflor yellow A (HSYA) on ischemic stroke. Eur. J. Pharmacol.818, 604–609. 10.1016/j.ejphar.2017.11.033

  • CrossRef
  • Google Scholar

• 28

  TangH.GuB. L.ZhouX. M. (2012). Study on the scores of blood stasis syndrome of acute ischemic stroke and its correlation with TOAST subtypes. Zhongguo Zhong Xi Yi Jie He Za Zhi32, 1500–1502.

  • Google Scholar

• 29

  VelchevaI.NikolovaG. (2008). Hemorheological disturbances and cognitive function in patients with cerebrovascular disease. Clin. Hemorheol. Microcirc.39, 397–402. 10.3233/ch-2008-1108

  • CrossRef
  • Google Scholar

• 30

  WangG.FangB.YuX.LiZ. (2018). Interpretation of 2018 guidelines for the early management of patients with acute ischemic stroke. Zhonghua Wei Zhong Bing Ji Jiu Yi Xue30, 289–295. 10.3760/cma.j.issn.2095-4352.2018.04.001

  • CrossRef
  • Google Scholar

• 31

  WangS.YuL.SunG.LiuY.HuW.LiuY.et al (2020a). Danhong injection protects hemorrhagic brain by increasing peroxiredoxin 1 in aged rats.Front. Pharmacol.11, 346. 10.3389/fphar.2020.00346

  • CrossRef
  • Google Scholar

• 32

  WangX.MaZ.FuZ.GaoS.YangL.JinY.et al (2016). Hydroxysafflor yellow A protects neurons from excitotoxic death through inhibition of NMDARs. ASN Neuro8, 1759091416642345. 10.1177/1759091416642345

  • CrossRef
  • Google Scholar

• 33

  WangY.WuH.HanZ.ShengH.WuY.WangY.et al (2022). Guhong injection promotes post-stroke functional recovery via attenuating cortical inflammation and apoptosis in subacute stage of ischemic stroke. Phytomedicine.99, 154034. 10.1016/j.phymed.2022.154034

  • CrossRef
  • Google Scholar

• 34

  WangY.ZhangL.PanY. J.FuW.HuangS. W.XuB.et al (2020b). Investigation of invigorating qi and activating blood circulation prescriptions in treating qi deficiency and blood stasis syndrome of ischemic stroke patients: Study protocol for a randomized controlled trial. Front. Pharmacol.11, 892. 10.3389/fphar.2020.00892

  • CrossRef
  • Google Scholar

• 35

  XueX.DengY.WangJ.ZhouM.LiaoL.WangC.et al (2021). Hydroxysafflor yellow A, a natural compound from Carthamus tinctorius L with good effect of alleviating atherosclerosis. Phytomedicine.91, 153694. 10.1016/j.phymed.2021.153694

  • CrossRef
  • Google Scholar

• 36

  ZengM.ZhouH.HeY.DuH.YinJ.HouY.et al (2021a). Danhong injection enhances the therapeutic effect of mannitol on hemispheric ischemic stroke by ameliorating blood-brain barrier disruption. Biomed. Pharmacother.142, 112048. 10.1016/j.biopha.2021.112048

  • CrossRef
  • Google Scholar

• 37

  ZengM.ZhouH.HeY.WangZ.ShaoC.YinJ.et al (2021b). Danhong injection alleviates cerebral ischemia/reperfusion injury by improving intracellular energy metabolism coupling in the ischemic penumbra. Biomed. Pharmacother.140, 111771. 10.1016/j.biopha.2021.111771

  • CrossRef
  • Google Scholar

• 38

  ZhangJ.ZhouR.CaoG.ZhangY.XuH.YangH. (2022). Guhong injection prevents ischemic stroke-induced neuro-inflammation and neuron loss through regulation of C5ar1. Front. Pharmacol.13, 818245. 10.3389/fphar.2022.818245

  • CrossRef
  • Google Scholar

• 39

  ZhouD.XieL.WangY.WuS.LiuF.ZhangS.et al (2020). Clinical efficacy of tonic traditional Chinese medicine injection on acute cerebral infarction: A bayesian network meta-analysis. Evid. Based. Complement. Altern. Med.2020, 8318792. 10.1155/2020/8318792

  • CrossRef
  • Google Scholar

• 40

  ZhouM.WangH.ZengX.YinP.ZhuJ.ChenW.et al (2019). Mortality, morbidity, and risk factors in China and its provinces, 1990-2017: A systematic analysis for the global burden of disease study 2017. Lancet394, 1145–1158. 10.1016/S0140-6736(19)30427-1

  • CrossRef
  • Google Scholar

• 41

  ZhuY.ZhangX.YouS.CaoX.WangX.GongW.et al (2020). Factors associated with pre-hospital delay and intravenous thrombolysis in China. J. Stroke Cerebrovasc. Dis.29, 104897. 10.1016/j.jstrokecerebrovasdis.2020.104897

  • CrossRef
  • Google Scholar