Prediction Model to Optimize Long-Term Antithrombotic Therapy Using Covert Vascular Brain Injury and Clinical Features

作者：Kaori Miwa, Kenta Tanaka, Masatoshi Koga, Kanta Tanaka, Yusuke Yakushiji, Makoto Sasaki, Kohsuke Kudo, Masayuki Shiozawa, Sohei Yoshimura, Masafumi Ihara, Shigeru Fujimoto, Haruhiko Hoshino, Kenji Kamiyama, Hiroyuki Kawano, Hikaru Nagasawa, Yoshinari Nagakane, Kazutoshi Nishiyama, Yoshiki Yagita, Shinichi Yoshimura, Teruyuki Hirano, Ḱazunori Toyoda, on behalf of the BAT2 Investigators · 发表于：Stroke · 年份：2025 · DOI：10.1161/strokeaha.125.050859 · 被引用次数：4 · 研究领域：Intracerebral and Subarachnoid Hemorrhage Research、Acute Ischemic Stroke Management、Cerebral Venous Sinus Thrombosis

BACKGROUND: Defining the risk of developing major bleeding, especially intracranial hemorrhage (ICH), or ischemic stroke (IS) in patients receiving antithrombotic therapy is crucial. Existing risk prediction tools would inadequately assess the net clinical benefit of antithrombotic therapy. We aimed to develop novel risk scores incorporating covert vascular brain injury to personalize the risk assessment of major bleeding, ICH, and IS in patients receiving antithrombotic therapy. METHODS: The prospective, multicenter, observational study (BAT2 [Bleeding With Antithrombotic Therapy Study-2]) enrolled patients receiving oral antiplatelets or anticoagulants from 52 hospitals across Japan between 2016 and 2019. Multimodal brain magnetic resonance imaging was performed at baseline under prespecified conditions to determine cerebral small vessel disease (white matter hyperintensity, cerebral microbleed, lacune, enlarged perivascular space, and cortical superficial siderosis), nonlacunar infarct, and intracranial artery disease with central reading. Risk scores, collectively termed the BAT2 scores, were developed separately to evaluate the comparative risks of (1) major bleeding, (2) ICH, and (3) IS based on covariates from Cox proportional hazards models and clinical relevance. Model performance was assessed with the Harrell C-index and calibration slope adjusted for optimism via bootstrapping. RESULTS: Of 5378 patients enrolled, 5250 were analyzed (mean age, 71±11 years, 33% women...