Study registration

This protocol for systematic review was carried out in full accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses for Protocols (PRISMA-P) guideline [21] and the PRISMA-extension statement for network meta-analysis [22]. We have registered this protocol on the International Prospective Register of Systematic Reviews (CRD42023459102). The PRISMA-P checklist is shown in S1 File.

Eligibility criteria

Data source and search strategy

We will systematically search PubMed, Embase, Cochrane Library, Chinese National Knowledge Infrastructure, VIP Database, Wan-fang Database, Chinese Biomedical Database. All of them will be retrieved from inception to 15, April 2024. The search terms were based on the combination of subject words and free words. Details of the search strategy for the Chinese and English databases are in the S2 File.

Study selection

Two reviewers will perform the literature selection independently according to the search strategy. We used EndNote 20 to manage the obtained literature, removing duplicates and then further reading the title, abstract, and full text to screen the literature that met the inclusion criteria. Any disagreement will be adjudicated by a third reviewer. The retrieval flow will be shown in a PRISMA flow chart as Fig 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. The details of the selection process.

https://doi.org/10.1371/journal.pone.0304174.g001

Data extraction

Two reviewers will independently extract relevant data from all studies ultimately included, following a pre-developed Excel data collection sheet. In case of missing study data, we will contact the authors via email to request complete data. All data will be cross-checked prior to data analysis and any differences will be discussed in the review panel to achieve a consistent result. The extracted data as described below:

1. Study characteristics: author, publication year, country, study design, study centers;
2. Participant characteristics: sample size, gender, age, diagnostic criteria, disease situation (disease course, infarct site, paralyzed limb), use of medications, treatment course, follow-up course;
3. Interventions and comparators details: type of electrical stimulation, duration, frequency, session, and main stimulation parameters;
4. Outcomes: all study measurement data, adverse events, time nodes, etc.

Risk of bias assessment

Two reviewers will evaluation the risk of bias in all included studies with the version 2 of the Cochrane risk-of-bias assessment tool (RoB 2) [24]. RoB 2 includes the following five projects: bias arising from the randomisation process, bias due to deviations from intended interventions, bias due to missing outcome data, bias in measurement of the outcome, bias in selection of the reported result. The overall predicted direction of bias for outcome as low risk of bias, some concerns and high risk of bias. In order to minimize bias caused by subjective factors of reviewers during the assessment process, detailed guidelines and training will be provided for reviewers before the formal start of the study. Meanwhile, any differences will be discussed and resolved in the review team.

Data synthesis

The study data were statistically analyzed using Stata 16.0 statistical software (version 17.0; Stata Corporation, College Station, Texas, USA). In the course of the study, the random-effects model was first used to conduct pairwise meta-analysis of the primary outcome, respectively discussing the clinical effects of various electrical stimulation therapy and control group. FMA as an index of continuous variable, we will use standardized mean differences with 95% confidence interval as pooled statistic. P-value < 0.05 was considered statistically significant. We will calculate the I² statistic and P-value to assess heterogeneity. Heterogeneity includes methodological heterogeneity and statistical heterogeneity, etc. If I² > 50%, we will consider the heterogeneity to be present, and we will further discuss the actual sources of heterogeneity, and analyze their potential impact on the findings in the discussion, especially the influence of different study characteristics.

For network meta-analysis, we will choose a random effects mode based on the frequency framework to analyze the direct and indirect evidence of all interventions. The outcome indicators for binary variables will be calculated using odds ratio, while the outcome indicators for continuous variables will be calculated using mean difference or standardized mean difference, both expressed as effect values and 95% confidence interval. First, we will use the "networkplot" command to plot the network relation graph to represent the quantitative relationships between studies and interventions. If the network relation graph can form a closed loop, then the node-splitting method is used to test the inconsistency. Second, data synthesis will be performed using a random-effects model of network meta-analysis to compare the relative efficacy of different electrical stimulation therapies. Finally, the "sucraprob" command was used to rank the efficacy of different interventions, and the cumulative probability graph was drawn. The surface under the cumulative ranking curve (SUCRA) was used to indicate the possibility of the pros and cons of the intervention. Higher SUCRA means that the intervention is more effective. The "netfunnel" command was used to make funnel plots, and publication bias and small sample evaluation were conducted for included studies. We will present the research results in a visual way.

Sensitivity analysis

We will conduct a sensitivity analysis to verify the robustness and reliability of the results by excluding studies that may have a high risk of bias to determine their potential impact on the results.

Grading the strength of evidence

We plan to use the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) framework to assess the strength of evidence for each outcome. The quality grade may be rated down for the following reasons: risk of bias, inconsistency, indirectness, imprecision, publication bias. The strength of evidence has four grades of high, moderate, low, and very low. High quality means that the true effect is close to the estimate of the effect and will be strongly recommended when making clinical decisions. Conversely, the very low quality indicates that the true effect is likely to be substantially different from the effect estimate [25].

Ethics and dissemination

The data needed for our study came from public databases, and the entire study did not involve patients and the public directly, so the ethical approval was not necessary. The final findings will be peer-reviewed and published in a journal.