Background/aim To clarify the nature of the relationship between novel oral anticoagulants (NOACs) and traditional anticoagulation in respect to intraocular bleeding.

Methods A comprehensive literature search up to October 2017 yielded 12 randomised controlled trials. Bayesian Markov chain Monte Carlo analysis was employed to investigate the relationship across multiple trials with varying NOACs. Random effects (informative priors) ORs were applied for the risk of intraocular bleeding due to various treatment measures. Mantel-Haenszel pairwise analyses were also performed. A total of 102 617 participants from 12 different randomised controlled trials. 11 746 received apixaban, 16 074 received dabigatran, 18 132 received edoxaban, 11 893 received rivaroxaban and 44 764 received warfarin.

Results Edoxaban was significantly associated with a reduced risk of intraocular bleeding in comparison to warfarin (OR 0.59; 95% CI 0.34 to 0.98). All other findings were non-significant; however, apixaban was the only NOAC to trend with an increased event rate against warfarin. The Bayesian Markov chain Monte Carlo modelling indicated that edoxaban had the greatest chance of producing the lowest rate of bleeding (surface under the cumulative ranking curve 0.8642). Pooled pairwise analysis supported the network analysis results favouring edoxaban against warfarin (OR 0.59; 95% CI 0.39 to 0.90; p=0.02) as well as subgroup analysis of low-dose edoxaban versus warfarin (OR 0.43; 95% CI 0.24 to 0.78).

Conclusion The analysis suggests that edoxaban may be the paramount agent in reducing intraocular bleeding rates. Given a paucity of reporting data for this rare event, future research and confirmation is strongly recommended.