Aims
The primary objective of this investigation is to evaluate the efficacy of acupuncture on delirium compared with placebo or any other interventions. Other outcomes are to assess its effect in different ways of acupoint stimulation, the incidence of adverse events, pain score, consumption of analgesics, number of cumulative delirious days, hemodynamics, as well as the length of hospital stay.
Design
The implementation of our systematic review will rigorously comply with the methodological recommendations of the Cochrane Handbook for Systematic Reviews and Meta analyses and the items of Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA-P) [18], and the checklist is available in Additional file 1. The protocol was prospectively registered on the homepage of the International Prospective Register of Systematic Reviews (PROSPERO).
Eligible criteria
Eligible criteria are determined based on PICOS (Population, Intervention, Comparator, Outcomes and Study Design) guidance, with detailed information presented in Table 1.
Information sources and search strategy
In order to obtain the most comprehensive literature, we will search PubMed, Embase, the Cochrane Library (Cochrane Central Register of Controlled Trials (CENTRAL)), Web of Science, Medline and the Cochrane Library, without language restrictions. We will split the search content into three core term categories (intervention (acupuncture, acupuncture needle, acupressure, transcutaneous electrical stimulation, moxibustion, acupressure, stimulating patches and pellets), outcome (delirium, postoperative delirium, POD), RCTs) to generate medical subject terms (MeSH) and free-text terms to retrieve literature. The search strategy will be verified by an experienced information specialist, which will be adapted for the aforementioned databases. Next, we will download all the search results and upload them to the literature management software, and then delete duplicate data automatically or manually.
Study selection
Two independent researchers will conduct the screening work in two stages in lines with the pre-established inclusion and exclusion criteria. In the first stage, the titles and/or abstracts will be scanned to remove items that do not meet the criteria, and then, qualified studies will go through full text review stages. Any disagreements arising at any step will be resolved instantly after negotiation or a third party.
Data abstraction
We will first establish a standardized data extraction accepted by the review team, in the form of excel. Data will be extracted by two members separately. Instead of directly compiling the extracted data, the data will be placed in a separate excel first. Two members will participate in reviewing all selections to reach an agreement. Disagreements will be settled by negotiation, and if it fails, the third member will be required to mediate. The study authors will be contacted for data clarification or insufficient data. Finally, all valid data will be aggregated into an individual master sheet. When the results are presented in the form of graphs, we will use an application (WebPlotDigitizer) [19] to extract the data.
We will extract the following main characteristics of all included studies:
-
Study characteristics: First author, publication year, country, total sample size, the number of patients analyzed in each group;
-
Patients characteristics: age, sex, prior chronic use of sedatives and dependence, whether the operation (type of surgery and anesthesia), type of population (geriatric or not);
-
Intervention and comparators: the acupoints, type, timing, the duration of treatment;
-
Outcomes: delirium (assessment tools, timing of evaluation), others (the incidence of adverse events, pain score, consumption of analgesics, number of cumulative delirious days, hemodynamics, the length of hospital stay)
Risk of bias assessment
Two researchers will be appointed to use Review Manager's (version 5.3) built-in Cochrane risk of bias assessment tool to separately evaluate the methodological quality of the final inclusion studies, which involved seven different domains:
-
random sequence generation (selection bias)
-
allocation concealment (selection bias)
-
blinding of participants and personnel (performance bias)
-
blinding of outcomes assessment (detection bias)
-
incomplete outcome data (attrition bias)
-
selective reporting (reporting bias)
-
other bias.
Each trial was classified as high risk or unclear or low risk after rigorous analysis. The result will finally be displayed graphically.
Data synthesis
Stata 12.0 (StataCorp LLC, College Station, USA) will be executed to pool and analyze the extracted data. Our primary outcome is the incidence of delirium, which can be determined in advance as dichotomous data, so when processing this data, the relative risk (RR) (or odds ratio (OR) if deemed appropriate) and a 95% confidence interval (CI) will represent the measure of effect size. But for other outcomes, effect size will be presented by specific data types, for continuous data, we will take the mean difference (MD) (or standardized mean difference (SMD) in case of different evaluation scale) with 95% confidence intervals [20]. Based on the random effects model, we use the I-squared statistics to quantify the significant heterogeneity. The heterogeneity will be identified considerable if I2 is more than 50% [21], and the effect model remains unchanged. Otherwise, we will consider replacing it with a fixed effects model. In addition, we will conduct sensitivity analysis and subgroup analysis to confirm the robustness of the results and attempt to find the sources of heterogeneity. Literature that can explain the significantly heterogeneity will be eliminated and narrated separately. Subgroup analysis will be carried out considering participants (gender, region, surgery), types of anesthesia (general anesthesia, spinal anesthesia or others), interventions (intervention methods (non-invasive or invasive), intervention timing), outcome (delirium assessment tool, assessment time). Meta-regression with multiple covariates will be performed, if possible. Moreover, we will conduct the trial sequential analysis (TSA) to minimize false positive results due to sparse sample size and random errors [22]. Publication bias will be assessed by funnel plots and Egger’s test (p-value > 0.5 means no significant asymmetry), if sufficient data are available. The strength of the evidence will be assessed in accordance with the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) guidelines [23], and the level of judgment is targeted is on the summary of the evidence.