External validation of risk scores for major bleeding in a population-based cohort of transient ischemic attack and ischemic stroke patients

Supplemental Digital Content is available in the text.
Background and Purpose— The S2TOP-BLEED score may help to identify patients at high risk of bleeding on antiplatelet drugs after a transient ischemic attack or ischemic stroke. The score was derived on trial populations, and its performance in a real-world setting is unknown. We aimed to externally validate the S2TOP-BLEED score for major bleeding in a population-based cohort and to compare its performance with other risk scores for bleeding. Methods— We studied risk of bleeding in 2072 patients with a transient ischemic attack or ischemic stroke on antiplatelet agents in the population-based OXVASC (Oxford Vascular Study) according to 3 scores: S2TOP-BLEED, REACH, and Intracranial-B2LEED3S. Performance was assessed with C statistics and calibration plots. Results— During 8302 patient-years of follow-up, 117 patients had a major bleed. The S2TOP-BLEED score showed a C statistic of 0.69 (95% confidence interval [CI], 0.64–0.73) and accurate calibration for 3-year risk of major bleeding. The S2TOP-BLEED score was much more predictive of fatal bleeding than nonmajor bleeding (C statistics 0.77; 95% CI, 0.69–0.85 and 0.50; 95% CI, 0.44–0.58). The REACH score had a C statistic of 0.63 (95% CI, 0.58–0.69) for major bleeding and the Intracranial-B2LEED3S score a C statistic of 0.60 (95% CI, 0.51–0.70) for intracranial bleeding. The ratio of ischemic events versus bleeds decreased across risk groups of bleeding from 6.6:1 in the low-risk group to 1.8:1 in the high-risk group. Conclusions— The S2TOP-BLEED score shows modest performance in a population-based cohort of patients with a transient ischemic attack or ischemic stroke. Although bleeding risks were associated with risks of ischemic events, risk stratification may still be useful to identify a subgroup of patients at particularly high risk of bleeding, in whom preventive measures are indicated.