Posttreatment Blood Pressure as a Key Predictor in a 5‐Year Stroke Prediction Model

ABSTRACT Evidence suggests that approximately 63.0%–84.2% of stroke survivors have hypertension, yet there is currently no stroke prediction tool specifically designed for individuals with hypertension. Using data from 20 702 hypertensive patients from the China Stroke Primary Prevention Trial (CSPPT), we developed a 5‐year stroke risk prediction model. This prospective study collected treated blood pressure every 3 months, resulting in 22 measurements over the study period. The model was internally validated using bootstrap resampling, and its predictive performance was assessed with the C‐index and calibration curves. We also developed a random forest model to rank the variable importance. The 5‐year stroke risk prediction model for hypertensive individuals includes 10 risk factors, ranked by importance as follows: average systolic blood pressure during treatment, age, average diastolic blood pressure during treatment, baseline systolic blood pressure, history of diabetes, baseline total cholesterol level, baseline folate level, self‐reported stress, smoking, and folic acid supplementation or not. The C statistic of the equation was 0.74 and there were no significant differences by gender or treatment group. Calibration plots indicate good internal consistency between observed and predicted 5‐year stroke risk. We also developed an online calculator to assist clinicians and patients (https://zhouziyi.shinyapps.io/CSPPT/). Our study indicates that for patients with hypertension, long‐term posttreatment blood pressure is the primary predictor of stroke risk. Trial Registration: The CSPPT (clinicaltrials.gov Identifier: NCT00794885).