Methods
Search Strategy
The research was in accordance with the Preferred Reporting Items for the Systematic Reviews and Meta-Analyses (PRISMA) guidelines.[16] We conducted electronic databases screening including Medline, Cochrane Library, PubMed, Embase, WEIPU database, CNKI database, WANFANG database from inception to January 2021. The controlled vocabulary came from Medical Subject Headings (MeSH) in PubMed and Chinese Medical Subject Headings (CMeSH) in SinoMed.
For example, "Sepsis" [MeSH], "DIC" [MeSH], "Heparin" [MeSH] and corresponding keywords were used for search with various combinations of the operators "AND" and "OR" in PubMed. The search was slightly adjusted according to the requirements of the various databases. The PubMed strategy details were presented in the Supplementary Table S1. At the same time, we manually searched the RCTs, meta-analyses, and systematic reviews for studies that were missed in the initial electronic search. There was no restriction on language or year of publication. The searching was duplicated and the last search update was January 2021. Furthermore, a third reviewer intervened whenever there was a disagreement.
Inclusion Criteria
Studies that were included met the following criteria:1. RCTs; 2. Studies that aimed to assess the clinical efficacy of UFH treatment on mortality, the incidence of bleeding complications and the coagulation indicators, such as the platelet (PLT), prothrombin time (PT) and activated partial thromboplastin time (APTT); 3. Participants in studies were diagnosed with sepsis, septic shock or DIC; 4. The administration of low-dose UFH, given at any route or frequency.
Exclusion Criteria
1.Quasi-randomized controlled trials, for example randomization according to hospital number; 2.Review, repeated literature reports, or animal studies; 3. Studies involving pediatric patients, pregnant or lactating patients; 4. Ex-transplant recipients, patients who had been treated with anticoagulant drugs 48 h ago and had coagulation disorders or history of abnormal coagulation; 5. Patients' PLT < 30 × 1011/L; 6. Patients with severe brain trauma, cerebral aneurysm, arteriovenous malformation and gastrointestinal bleeding history; 7. Renal failure required hemodialysis (hemodialysis) or hemofiltration (hemofiltration); 8. Both the experimental and control group received UFH.
Data Extraction
Two researchers independently extracted the data from each article that met the inclusion criteria but not exclusion criteria. If the opinion of two reviewers conflicted in the process, another reviewer evaluated the studies independently. We then finalized the decisions through group discussion. The following data were extracted: the first author's name, year of publication, number of study sites, number of patients, population, age, Acute Physiology and Chronic Health Evaluation II (APACHE II) score, intervention (dose and duration), control treatment, the 28 d mortality and incidence of bleeding complications.
Outcomes
The primary outcome was 28 d mortality. We made reference to APACHE II score to perform subgroup analysis. Also, PLT, PT and APTT were regarded as the coagulation indicators. Multiple organ dysfunction syndrome (MODS) incidence, length of stay (LOS) and the duration of ventilation in ICU were recorded as the secondary outcomes. We graphically displayed the outcomes by forest plots and visually inspected the potential publication bias with a funnel plot.[17]
Assessment of Risk of Bias
The Cochrane Collaboration Risk Of Bias tool (ROB) was utilized to assess the quality of all analyzed trials by two authors.[18] Disagreements between the reviewers were resolved by negotiating with another author. ROB estimated the selection bias by random sequence generation and allocation concealment, performance bias by blinding of participants and personnel, detection bias by blinding of outcome assessment, attrition bias by incomplete outcome data, reporting bias by selective reporting and other potential sources of bias.[18] Each domain was referred to as "low risk," "high risk," or "unclear risk" which was identified depended on the researcher' response.[18] The overall risk of bias for the result was the least favourable assessment across the domains of bias.[18]
Statistical Analysis
The meta-analysis was performed using the Review Manager (RevMan) version 5.3 software. Reference to Cochrane Handbook of Systematic Reviews, 28 d Mortality and bleeding complication (Binary variables) were expressed as relative risk (RR) with 95% confidence intervals (CIs), a weighted pooled RR was calculated using the Mantel-Haenszel method.[19] The continuous variables were expressed as mean differences (MD) with 95% CIs, for example, MODS incidence and LOS in ICU.[20] Jaimes et al' recorded the APACHE II score as median and interquartile range (IQR), We calculated its mean and standard deviation according to the sample size with a calculator.[20]
I2 test was used to measure statistical heterogeneity. When I2 = 0, there was no heterogeneity. When the I2 < 40% or I2 between 30 and 60%, there was considered as low or moderate statistical heterogeneity in the studies, the fixed effect model was used for analysis.[19] Random effects model was used to analyze studies which had an I2 > 50%.[19] All results were summarized in forest plots with both point estimates and alpha = 0.05 displayed.
Evidence Grade
Two researchers graded the evidence for each outcome based on the following six domains: study design, risk of bias, inconsistency, indirectness, imprecision, and publication bias. If there were discrepancies, another author would take part in this discussion. The strength of evidence could be classified as "high," "moderate," or "low".[21]
BMC Anesthesiol. 2022;22(28) © 2022 BioMed Central, Ltd.
