Risk Stratification Tool for Ischemic Stroke: A Risk Assessment Model Based on Traditional Risk Factors Combined With White Matter Lesions and Retinal Vascular Caliber

Abstract

Objective: This study aims to establish a risk assessment model based on traditional risk factors combined with the Fazekas classification of white matter lesions and retinal vascular caliber for screening the patients at high risk of ischemic stroke.Methods: This study included 296 patients (128 cases of ischemic stroke and 168 cases in the normal control group). The basic data of the patients were collected. Color fundus photography was performed after pupil dilation, and the retinal vascular caliber was measured using semiautomated vascular measurement software (IVAN Software, Sydney, Australia). The severity of white matter lesions (WML) on cranial nuclear magnetic fluid-attenuated inversion recovery images were assessed using the Fazekas scale. Moreover, logistic regression analysis was used to establish different risk assessment models for ischemic stroke. The effects of models were evaluated through the receiver operating characteristic (ROC) curve and the Delong test compared area under the curve.Results: The sensitivity and specificity of models 1 (the traditional risk factor model), 2 (the retinal vascular caliber model), 3 (the WML model), and 4 (the combined the traditional risk factor, WML and central retinal artery equivalent (CRAE) model) were 71 and 55%, 48 and 71%, 49 and 67%, and 68 and 68.5% with areas under the curve of 0.658, 0.586, 0.601, and 0.708, respectively. The area under the receiver operating characteristic curve in models 1, 2, 3, and 4 showed statistically significant differences. Moreover, no statistical significance exists in the pairwise comparison of other models.Conclusion: The risk assessment model of ischemic stroke combined with Fazekas grade of WML and CRAE is superior to the traditional risk factor and the single-index model. This model is helpful for risk stratification of high-risk stroke patients.