Biomarker-Based Prediction of Ischemic Stroke in Patients With H-type Hypertension

Hypertension combined with hyperhomocysteinemia significantly raises the risk of ischemic stroke. Our study aimed to develop and validate a biomarker-based prediction model for ischemic stroke in H-type hypertension patients. We retrospectively included 3,305 patients in the development cohort, and externally validated in 103 patients from another cohort. Logistic regression, LASSO regression, and best subset selection analysis were used to assess the contribution of variables to ischemic stroke, and models were derived using four machine learning algorithms. Area Under Curve (AUC), calibration plot and decision-curve analysis (DCA) respectively evaluated the discrimination and calibration of four models, then external validation and visualization of the best-performing model. There were 1,415 and 42 patients with ischemic stroke in the development and validation cohorts. The final model included 8 predictors: age, antihypertensive therapy, biomarkers (serum magnesium, serum potassium, proteinuria and hypersensitive C-reactive protein), and comorbidities (atrial fibrillation and hyperlipidemia). The optimal model, named A ₂ BC ischemic stroke model, showed good discrimination and calibration ability for ischemic stroke with AUC of 0.91 and 0.87 in the internal and external validation cohorts. The A ₂ BC ischemic stroke model had satisfactory predictive performances to assist clinicians in accurately identifying the risk of ischemic stroke for patients with H-type hypertension.