2.1. Search strategy and study selection
We registered our review in PROSPERO, and registration ID is CRD42021225236. We performed a search on PubMed, EMBASE, Cochrane Library, Web of Science, and ClinicalTrials.gov databases for eligible studies. We used OpenGrey and National Technical Information Service databases to search relevant grey literature. The text words or MeSH were “randomized controlled trail” “intracranial pressure”, “elevated intracranial pressure”, “hyperosmolar agent”, “hypertonic saline”, “mannitol” and other synonymous free text words and phrases were substituted for searching comprehensively. We did not restrict studies based on language and status of publication. And the published date was restricted from 2000 to now. We also searched previous published similar meta-analysis, and screened the included studies. The search was completed by two investigators independently. And all disagreements were determined by a third investigator. Then there were three independent investigators reviewed all references and screened out eligible studies which conformed to the inclusion and exclusion criteria. And all disagreements were determined by a fourth investigator.
2.2. Inclusion criteria :
The included literatures should meet the following items: Patients: (1) Adult (≥ 16 years old, and of both sexes), (2) episodes of elevated ICP occurred when used osmotic agents, (3) ICP monitoring (undergoing quantitative ICP measurement). Interventions: treatments included both HTS and mannitol. Outcomes: the available quantitative data of ICP. Study: randomized controlled trails, full text available.
Exclusion criteria :
Studies met the following criteria were excluded: (1) No threshold value of ICP when osmotic agents were used. (2) Patients with liver or renal failure, cardiac dysfunction, hypovolemic shock, or multiple organ failure. (3) Qualitative trials, which had no exact ICP values. (4) Prehospital studies. (5) Animal studies, retrospective studies, cohort studies, case report, meta-analysis, and reviews.
Data extraction
To reduce bias, two independent authors extracted data from included studies. And another author checked the consistencies of the two sets of data. If there were disagreements, the final decisions were made by discussion. The basic information included the followings: first author’s name, publication date, study design, country, sample size, interventions. Furthermore, the screened clinical outcomes of interest as followings: (1) ICP, (2) CPP, (3) heart rate (HR), (4) mean arterial pressure (MAP), (5) serum sodium, (6) serum osmolarity, (7) hematocrit (HCT). Several interested data were missing, and we had contacted the corresponding authors to get the information.
Risk of bias
Three authors assessed risk of bias of included studies independently. Another author solved inconsistencies, if necessary. The Cochrane tool framework was used to assess risk of bias (9). Mainly included the following items: (1) random sequence generation (selection); (2) allocation concealment (selection bias); (3) blinding of participants and personal (performance bias); (4) blinding of outcomes assessment (detection bias); (5) incomplete outcome data (attrition bias); (6) selective reporting (reporting bias); (7) Other bias. For each item, we assessed the risk of bias as “low risk”, “high risk” or “unclear”. And we generated a risk of bias summary figure upon completion of these assessment.
Statistical analysis
Although data in different articles were defined variably, we changed them in a unified form by the Cochrane Handbook (10). Not every extracted outcomes could be analyzed meaningfully, we only performed meta-analyses if outcomes were investigated by at least 3 RCTs. The continuous variables were pooled using the mean differences (MD) and 95% confidence interval (Cl). The heterogeneity of the studies was assessed by Q test (P < 0.10 as regards for significant heterogeneity) and I2 statistic (I2= 0%-25%, no heterogeneity; I2 = 25%-50%, moderate heterogeneity; I2 = 50%-75%, large heterogeneity; I2 = 75%-100% extreme heterogeneity) (11). A random-effect model was applied when I2 > 50%, otherwise fix-effect model was applied. We performed a sensitivity analysis by excluding each study in turn for outcomes to investigate the potential source of heterogeneity and effect of each study on pooled results. And publication bias was estimated by a contour-enhanced funnel plot and Egger’s test. All data were analyzing by Review Manager 5.3 and STATA 16.
Characteristics and conclusions of similar systematic reviews and meta-analyses
We searched and selected similar systematic reviews and meta-analyses on PubMed, EMBASE, Cochrane Library, Web of Science, and ClinicalTrials.gov databases. And key information and conclusions were extracted for further analysis. For methodological quality evaluation, the AMSTAR 2 (A MeaSurement Tool to Assess systematic Reviews 2) assessment tool was used. AMSTAR 2 consists of 16 items (7 critical domains: item 2, 4, 7, 9, 11, 13, 15), graded as “Yes,” “Partial Yes,” and “No” (12). Rating overall confidence in the results of reviews is as follows: “Critically low”, “Low”, “Moderate” and “High” (12). Three independent authors evaluated the 16 items and rated all reviews. And the inconsistencies were discussed by all three authors.
Results