Meta-Analysis of Bleeding Scores Performance for Acute Coronary Syndrome.

Background: Bleeding is a common and frequently devastating complication in acute coronary syndrome (ACS). It is critical to evaluate in the current era of ACS management involving invasive strategies and potent anti-thrombotics. Risk models remain under-utilised in this setting but may guide the choice and duration of therapy. We compared their performances for predicting bleeding in ACS patients in this meta-analysis.

Methods: Medline, EMBASE, Cochrane and Scopus were searched for relevant articles from 1980 to 31 December 2017 assessing external validation of risk scores for bleeding after ACS. Two (2) authors independently reviewed the searched studies for eligibility, followed by pooled analyses using random effects models.

Results: Amongst 1,843 articles searched, 73 full-texts were reviewed and 17 studies totalling 18,155 patients were included for analysis. C-statistics (95% confidence interval) for predicting in-hospital major bleeding by risk model were Can Rapid risk stratification of Unstable angina patients Suppress ADverse outcomes with Early implementation of the ACC/AHA guidelines (CRUSADE) 0.714 (0.659-0.779), Acute Catheterization and Urgent Intervention Triage strategY (ACUITY) 0.711 (0.626-0.797), Acute Coronary Treatment and Intervention Outcomes Network (ACTION) 0.767 (0.737-0.797), Global Registry of Acute Coronary Events (GRACE) 0.689 (0.473-0.905) and HAS-BLED 0.636 (0.460-0.812). CRUSADE also predicted bleeding during medium-term follow-up c=0.704 (0.644-0.765). It performed better for radial versus femoral access (c=0.826 and 0.734), invasive versus non-invasive strategy (c=0.752 and 0.625) and similarly for ST elevation myocardial infarction (STEMI) and non-ST elevation myocardial infarction (NSTEMI) (c=0.791 and 0.760). Heterogeneities of studies and paucity of studies assessing risk scores beyond CRUSADE were important limitations.

Conclusions: Acute coronary syndrome-specific bleeding scores had moderate discrimination for bleeding, while the GRACE and HAS-BLED scores could not. The ACTION score had the highest pooled c-statistic, while the CRUSADE score was the most widely studied, and also performed better for invasive strategy and radial access subgroups.