Premium
Evaluation of the predictive performance of bleeding risk scores in patients with non‐valvular atrial fibrillation on oral anticoagulants
Author(s) -
Beshir S. A.,
Aziz Z.,
Yap L. B.,
Chee K. H.,
Lo Y. L.
Publication year - 2018
Publication title -
journal of clinical pharmacy and therapeutics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.622
H-Index - 73
eISSN - 1365-2710
pISSN - 0269-4727
DOI - 10.1111/jcpt.12634
Subject(s) - medicine , atrial fibrillation , dabigatran , rivaroxaban , warfarin , logistic regression , population , cohort , gastrointestinal bleeding , major bleeding , anticoagulant , emergency medicine , environmental health
Summary What is known and objective Bleeding risk scores ( BRS s) aid in the assessment of oral anticoagulant‐related bleeding risk in patients with atrial fibrillation. Ideally, the applicability of a BRS needs to be assessed, prior to its routine use in a population other than the original derivation cohort. Therefore, we evaluated the performance of 6 established BRS s to predict major or clinically relevant bleeding ( CRB ) events associated with the use of oral anticoagulant ( OAC ) among Malaysian patients. Methods The pharmacy supply database and the medical records of patients with non‐valvular atrial fibrillation ( NVAF ) receiving warfarin, dabigatran or rivaroxaban at two tertiary hospitals were reviewed. Patients who experienced an OAC ‐associated major or CRB event within 12 months of follow‐up, or who have received OAC therapy for at least 1 year, were identified. The BRS s were fitted separately into patient data. The discrimination and the calibration of these BRS s as well as the factors associated with bleeding events were then assessed. Results A total of 1017 patients with at least 1‐year follow‐up period, or those who developed a bleeding event within 1 year of OAC use, were recruited. Of which, 23 patients experienced a first major bleeding event, whereas 76 patients, a first CRB event. Multivariate logistic regression results show that age of 75 or older, prior bleeding and male gender are associated with major bleeding events. On the other hand, prior gastrointestinal bleeding, a haematocrit value of less than 30% and renal impairment are independent predictors of CRB events. All the BRS s show a satisfactory calibration for major and CRB events. Among these BRS s, only HEMORR 2 HAGES ( C ‐statistic = 0.71, 95% CI 0.60‐0.82, P < .001) and ATRIA score ( C ‐statistic = 0.70, 95% CI 0.58‐0.82, P < .001) show acceptable discrimination performance for major bleeding events. All the 6 BRS s, however, lack acceptable predictive performance for CRB events. What is new and conclusion To the best of our knowledge, this is the first evaluation study of the predictive performance of these 6 BRS s on clinically relevant bleeding events applied to the same cohort consisting of mainly Asian novel oral anticoagulant users. These BRS s show poor to acceptable predictive performance on OAC ‐induced major or CRB events. An improvement in the existing BRS s for OAC users is warranted.