Risk of bleeding with oral anticoagulants: an updated systematic review and performance analysis of clinical prediction rules

This study aimed to determine whether progress in developing bleeding risk estimation tools for patients on oral anticoagulant therapy has been made since 2006 when we last systematically reviewed this topic, and to refresh previously published quantitative evaluations of the clinical prediction rules (CPRs) available for estimating bleeding risk in patients on oral anticoagulant therapy. A systematic review of English language literature published since December 2006 was conducted when our previous systematic review ended. Studies were analyzed for predictive performance using likelihood ratios. Six studies detailed CPRs used to assess risk of bleeding prior to commencing warfarin therapy, four of which were included in the analysis. Three studies evaluated new CPRs ("RIETE" and "HAS-BLED"). One of the studies was a further validation of the modified outpatient bleeding risk index (mOBRI) in patients with atrial fibrillation. Individual trial and pooled analyses using likelihood ratios for mOBRI and HAS-BLED showed they have weak predictive accuracy. A RIETE score of 0 point was moderately predictive of the absence of major bleeding. None of the CPRs exhibited sufficient predictive accuracy or had sufficient validation to be recommended for routine use in practice. None of the available CPRs exhibit sufficient predictive accuracy or have trials evaluating the impact of their use on patient outcomes. Hence, no existing oral anticoagulation major bleeding CPR can be recommended for routine use in practice at present.