Registration
We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) to report our systematic review and meta-analysis[18] (Supplementary Appendix A). The protocol of our study was registered in International Prospective Register of Systematic Reviews (PROSPERO).
Search strategy
Two reviews (YZ) performed a comprehensive search on the Cochrane Library, PubMed, CINAHL, Web of Science, EMBASE, J-STAGE, KoreaMed, and CNIK from inception to 1 May 2022 with no restrictions on language or year of publication. Eligible studies were identified using the following key terms: nafamostat mesilate, nafamostat mesylate, NM, continuous renal replacement therapy, CRRT, continuous RRT, CVVH, continuous venovenous hemofiltration, CVVHD, continuous venovenous hemodialysis, CVVHDF, continuous venovenous hemodiafiltration, SCUF, and slow continuous ultrafiltration (Supplementary Appendix B). We manually searched the references of the collected articles and systematic reviews and also searched ongoing and unpublished trials using the clinicaltrials.gov databases to obtain additional eligible papers.
Eligibility criteria
Studies meeting the following criteria were candidates for inclusion: (1) studies that included patients with increased bleeding risk who underwent CRRT; (2) the conventional anticoagulant strategy was compared with NM. Studies with any of the following conditions were excluded: (1) full text was not available; (2) studies that did not analyse the safety and efficacy of NM in patients with high bleeding risk; (3) studies with insufficient or incorrect data; (4) the following article styles: review articles, case reports, letters, editorials, conference abstracts, and comments.
Study identification and data extraction
Two reviewers (YWL, QJL) independently screened the titles and abstracts for eligibility. Articles were retrieved in full upon request from reviewers. Then, reviewers independently screened the full texts and resolved disagreements through discussion. If they could not reach an agreement, another author (YYZ) was consulted and a decision was made by a majority vote.
Data were extracted independently by two authors (LC, PW) using predetermined forms. The following data were collected: (1) characteristics of the included studies: first author, publication year, study design, setting, inclusion and exclusion criteria, sample size, population characteristics, interventions, and endpoints; (2) information related to the CRRT: type of CRRT, CRRT machine, NM doses, filter, blood flow, replacement/dialysate fluid, pre/post-filter dilution, vascular access, and catheter size.
Quality assessment
Two reviewers (XYG, YYZ) conducted risk of bias assessment using the Cochrane Risk of Bias methodology[19] for RCTs and a modified version of the Newcastle-Ottawa Scale for observational studies[20] (Supplementary Appendix F). The two assessment tools were all developed by the Cochrane Collaboration. In addition, we used the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) approach to classify the certainty of evidence into high, moderate, low or very low for each outcome[21].
Data synthesis and statistical analysis
Meta-analysis was performed using Review Manager (RevMan, Version 5.3: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014, Copenhagen, Denmark). Heterogeneity among studies was assessed using the chi-square test, and Ι2 values were used to determine heterogeneity across studies. A random or fixed effects model was used to calculate the pooled effect sizes and corresponding 95% confidence intervals (CI) based on the heterogeneity. If the Ι2 ≦50%, which represented homogeneity, fixed-effects models were selected. If the Ι2 >50%, which indicated substantial heterogeneity of effects, random-effects models were applied. For continuous data, the mean difference (MD) and 95% CI were assessed for outcomes pooled and for dichotomous outcomes, the odds ratio (OR) with 95% CI was used in accordance with intent-to-treat principles. A forest plot was generated to represent the meta-analysis results. Subgroup analysis was conducted according to the study design. To test the robustness of the pooled estimates, we evaluated whether fixed-effects models and random-effects would bring about the same outcome. The stability of the results was also performed by sequentially excluding the included studies[22]. If more than 10 studies were included in the analysis of outcomes, the funnel plots were constructed to identify publication bias by Egger tests (with p < 0.05 considered significant) [23].