To systematically compare the effects of various antithrombotic strategies on prespecified outcomes including 28-day all-cause mortality (primary outcome), major thrombotic events and major bleeding events (secondary outcomes) in adult COVID-19 patients.

Systematic review and Bayesian network meta-analysis (NMA).

PubMed, Web of Science, Embase, Cochrane Library and ClinicalTrials.gov up to February 2024.

We included randomised controlled trials (RCTs; published in English) comparing different antithrombotic strategies (eg, anticoagulants, antiplatelet (AP) agents, fibrinolytics or combinations) in adults (aged≥18 years) with laboratory-confirmed SARS-CoV-2 infection. Eligible trials had at least one active antithrombotic arm versus another strategy or standard care.

Two reviewers independently extracted data using a standardised form; disagreements were resolved by consensus or third-party adjudication. Bayesian NMA was performed using Markov chain Monte Carlo methods with random/fixed effects models selected by the deviance information criterion. The risk of bias (RoB) was assessed using the Cochrane Collaboration’s tool. The confidence in NMA framework was used to assess the quality of evidence.

35 RCTs that randomly assigned 39 949 participants were included in the main analysis. Primary outcome: evidence of low to moderate certainty suggested that, compared with standard of care (SoC), both prophylactic-dose anticoagulation (PA) (risk ratio (RR) 0.71, 95% credible interval (CrI) 0.44 to 0.99) and therapeutic-dose anticoagulation (TA; RR 0.65, 95% CrI 0.38 to 0.94) reduced the 28-day all-cause mortality. Secondary outcomes: TA (RR 0.19, 95% CrI 0.09 to 0.31), TA+AP (RR 0.27, 95% CrI 0.05 to 0.95), PA (RR 0.33, 95% CrI 0.18 to 0.53) and AP+PA (RR 0.52, 95% CrI 0.25 to 0.94) were effective in reducing major thrombotic events. AP was associated with an increased risk of major bleeding events (RR 2.27, 95% CrI 1.01 to 5.07). Subgroup analyses by hospitalisation status showed that PA significantly reduced 28-day mortality versus SoC (RR 0.52, 95% CrI 0.26 to 0.90) for non-hospitalised patients, whereas no strategies showed significant benefit in hospitalised patients. Subgroup analysis based on severity of hospitalised patients indicated that TA was more favourable than PA in decreasing the 28-day mortality in non-critically ill patients (fixed-effect model: RR 0.75, 95% CI 0.61 to 0.91; random-effect model: RR 0.71, 95% CI 0.48 to 1.05), but for critically ill patients, all antithrombotic strategies showed no significant difference.

Our NMA indicates that both PA and TA reduced the 28-day all-cause mortality of adult COVID-19 patients. However, subgroup analyses revealed substantial heterogeneity, and the benefit may differ across hospitalisation status and disease severity.

CRD42022355213.