1. Literature resources
1.1 Criteria for inclusion
1.1.1 Research type: The types of included literature must be based on RCT, regardless of whether there is a blind method.
1.1.2 Research objects: The subjects met the diagnostic criteria of ischemic stroke formulated by the Fourth National Cerebrovascular Disease Conference, and were confirmed by brain computed tomography(CT)or magnetic resonance imaging(MRI). There was no restriction on basic clinical conditions such as age, sex and severity of the disease.
1.1.3 Intervention measure: RCTs of PNS combined with conventional WM treatment and conventional WM alone were included. Patients in the control group were treated with conventional WM treatment, including mannitol, atorvastatin calcium, aspirin, clopidogrel, citicoline, etc. Patients in the experimental group were treated with PNS on the basis of control group treatment, including Xuesaitong and Xueshuantong. The drug dosage forms included injection and capsule.
1.1.4 Outcome indicator: All literatures used the total effective rate as the outcome measurement indicator, and the total effective rate(%)=(number of basically recovered patients + number of significantly improved patients + number of improved patients)/total number of patients×100%. Changes in neurological deficit scores before and after treatment were assessed by the NIHSS. Specific indicators were, basic recovery: NIHSS decreased by more than 90%; significant improvement: NIHSS decreased by 46%-91%; improvement: NIHSS decreased by 18%-46%; no change: NIHSS changed within 18%; deterioration: NIHSS increased by more than 18% or death.
1.2 Criteria for exclusion
1.2.1 Duplicate publications; 1.2.2 The type of study was not clearly stated or the type of study was not a RCT; 1.2.3 The experimental group or control group was interfered by other treatment measures; 1.2.4 The data is incomplete and full text cannot be obtained.
2. Search strategy of literature
Search PubMed, Web of Science, Cochrane Library, Medline, Embase, EBSCO, China National Knowledge Infrastructure Database(CNKI), Wanfang Database, China Science and Technology Journal Database(VIP), Chinese Biomedical Literature Database(CBM). All articles published up to June 30, 2024 were searched.
For the English databases, we used the combination of subject headings and text words to search, and the search subject headings were "PNS" and "stroke". For the Chinese databases, we used "Sanqi" or "panax notoginseng" or "PNS" or "panax notoginseng total soap" or "panax notoginseng extract" for preliminary search, on this basis, the second search was conducted by using "stroke" or "ischemic stroke" or "cerebrovascular accident" or "acute stroke" or "acute cerebrovascular accident" or "stroke acute stage". The search process did not limit the type of literature and publication time, nor did it limit the language.
After the search strategy was formulated, two investigators independently conducted the literature search. If there was a dispute in the process, they discussed it together and resolved it. If no agreement could be reached, the third investigator intervened.
3. Data extraction and quality evaluation of literatures
The following data were extracted from the included literatures: 1. Basic elements of the literatures(including the first author and the year of publication). 2. Basic information of patients and whether there were differences between the baseline data of the two groups. 3. Intervention measures. 4. Evaluation indicators of treatment outcomes. 5. Occurrence of ADR.
The methodological quality of the included RCTs were evaluated using the Cochrane Risk Bias Assessment Tool, which included: 1. Randomization sequence generation. 2. Allocation scheme concealment. 3. Blinding of subjects and investigators. 4. Completeness of results data. 5. Whether study results were selectively reported. 6. Whether there were other sources of bias(such as selective reporting). Each item was classified into low risk, unclear and high risk bias. The evaluation was done independently by two investigators and checked against each other. Any disagreements regarding data extraction and quality evaluation were resolved through discussion, or if agreement could not be reached, by the third investigator.
4. Data statistics
4.1 Selection of efficacy indicators
At present, meta-analysis can only carry out combined analysis of continuous variables and binary variables. According to clinical practice, basic recovery, significant improve and improve are combined as effective, and the rest are ineffective. The effective rates among the study groups are compared and combined.
4.2 Meta-analysis
RevMan5.3 software was used for data synthesis in this literature. Relative risk(RR)was used as analysis statistic for binary variables, mean difference(MD)was used as analysis statistic for continuous variables, and 95% confidence interval(95%CI)was used for each effect size. P<0.05 was considered that the difference between experimental group and control group was statistically significant.
4.3 Heterogeneity test
Forest plots generated by RevMan5.3 software were used for heterogeneity test. The statistics of heterogeneity test within forest plots obeyed distribution. According to α=0.1 test level, if P<0.1 was satisfied, heterogeneity existed among studies. At the same time, I2 was used for quantitative analysis of heterogeneity, and its significance level was set at 50%, that is, I2>50% was considered as large heterogeneity between studies, and fixed effect model was used for pooled effect size analysis between studies without heterogeneity, otherwise random effect model was used.
4.4 Sensitivity analysis
The stability of the results was tested by a sensitivity analysis between the random effects model and the fixed effects model.
4.5 Detection of publication bias
If the number of RCTs included is greater than 10, funnel plots will be drawn to assess potential publication bias.