2 Design/Setting/Main Outcome
2.1 Literature search
The systematic review was done in accordance with the PRISMA 202014 statement. Two authors (PS and TW) searched independently on PubMed and the Cochrane database on April 25th 2021 using the following criteria: (epistaxis AND anticoagulation) OR (epistaxis AND DOAC). The search was restricted to the years following the FDA approval15 of the first DOAC (2009-2021). Disagreements were settled through consensus after thorough analysis. Additionally, the references of reviewed papers were also subject to analysis. No automatization tools were used in the process.
2.2 Study selection, bias assessment and data retrieval
The inclusion criterion was two-armed studies (DOAC vs. VKA) on consecutive epistaxis patients with a minimum of 20 patients in each group. The exclusion criteria were absence of a thorough comparison between DOAC and VKA groups, group size of <20 patients, duplicate studies, case reports, reviews, comments, animal studies, and letters to the editor. Data extraction was performed by two authors (PS and TW) independently. All studies were assessed for bias using the MINORS16 score for comparative non-randomized studies. Discrepancies were settled after a consensus between all authors had been reached.
2.3 Outcomes
The following outcomes were extracted: total number of epistaxis patients taking DOACs and VKAs, admission rates, days in hospital, posterior epistaxis, interventions needed, transfusions, recurrence rates, and haemoglobin values.
2.4 Statistical analysis
The meta-analysis was performed using the MedCalc Statistical Software version 17.8.6 (MedCalc Software bvba, Ostend, Belgium; http://www.medcalc.org; 2017). A standardized mean difference (SMD) with a 95% confidence interval (CI) and a pooled standard deviation (SE) calculated according to Hedges’ g was used to compare the days in hospital and haemoglobin values. Odds ratios (OR) with 95% CI according to the Mantel-Haenszel method were used to compare the rates of admission, posterior epistaxis, interventions, transfusions, and recurrence. Heterogeneity was accessed according to I2statistics. When the I2 statistics did not reach statistical significance, the fixed effects model was used; otherwise the random effects model was applied. Forest and funnel plots were created for each investigated result. A two sided P value <.05 was considered statistically significant.