Comparative Analysis Of Risk Stratification Of Acute Stage Of Ischemic Cerebrovascular Disease With Relapse

Background:Cerebrovascular disease has the characteristics of high incidence, high recurrence rate, high morbidity and high mortality. Recurrence of cerebrovascular disease can be prevented by doing recurrence risk assessment and risk factor intervention early, it has a very important significance for the prevention and treatment of cerebrovascular disease. At present, the ESRS and the ABCD2score are the instruments most used to stratifying recurrence risk of stroke. The ABCD2score was mainly used to assess whether TIA patients have an early stroke recurrence risk, whether TIA patients need an emergency treatment or to be admitted to hospital. The ESRS was mainly used to predict1-year risk of recurrent stroke after ischemic stroke. Although the ABCD2score and ESRS were simple clinical scores, and mainly used clinical symptoms and signs and other risk factors for risk prediction, the ABCD2score and ESRS did not include the laboratory or vascular imaging risk factors which had important influence on the prognosis of stroke, which may affect the predictive power of the ABCD2score and ESRS for stroke recurrence. We developed a new stroke risk assessment model based on the ABCD2score and ESRS. The new model incorporated the clinical symptoms and signs, laboratory and vascular imaging risk factors. We hope the new model can increase the predictability of the ABCD2score and ESRS for recurrent stroke, and the new model can guide the work of the secondary prevention of ischemic cerebrovascular disease.Objectives:(Ⅰ) Based on the clinical characteristics, imaging and laboratory risk factors of patients with ischemic cerebrovascular disease, this study developed a Stroke Risk Score (SRS).(Ⅱ) We assessed the predictability of the SRS scoring model for recurrent stroke of patients with ischemic cerebrovascular disease, and compared the SRS scoring model with ESRS and the ABCD2score.Methods:(Ⅰ) Development of the SRS scoring model:(Ⅰ) The potential components of the original model of SRS included age, blood pressure, blood glucose, Low-density lipoprotein cholesterol, homocysteine, C-reactive protein, atrial fibrillation, carotid imaging, previous transient ischemic attack or ischemic stroke, previous myocardial infarction, peripheral arterial disease, and so on.(2) Each variable in relation to the2-year risk of recurrent stroke was determined by univariate Cox regression. First, we included any variable that was a univariate predictor of the2-year risk of recurrent stroke with a significance of P≤0.10into the model A. Any of the predefined independent predictors not included in the model A, we studied the predictive power of them individually, and included the independent predictors that can overall improve the predictive power of each model with a significance of P<0.05. Through the step-by-step analysis, we finally determined the components of the SRS scoring model and the best weighting scheme for each independent predictors of the SRS scoring model.(Ⅱ) Assessment of predictability, groups:300patients of derivation group were hospitalized patients with ischemic cerebrovascular disease from1st September2009to31st August2010,2-year follow-up;315patients of validation group were hospitalized patients with ischemic cerebrovascular disease from1st September2010to31st August2011,1-year follow-up; the pooled group included615patients of derivation and validation groups,615patients were hospitalized from1st September2009to31st August2011; the TIA group included65TIA patients of derivation and validation groups, patients were hospitalized from1st September2009to31st August2011. Analysis:(1) Calculating the C-statistics of each scale to assess the predictive value of each scale;(2) Evaluating the sensitivity and specificity of each scale to predict patients of the pooled group into risk group with receiver operating characteristic curve, to compare the distinction ability of each scale to distinguish patients with high risk of recurrent stroke;(3) Assessing the survival curves of low risk or high risk groups of each scale with Kaplan-Meier curve, to compare survival rates without stroke of low risk or high risk groups;(4) Investigating the trend of observed risk compared with predictive risk.Results:(I) Development of the SRS scoring model:The components of SRS scoring model included age, diabetes mellitus, low-density lipoprotein cholesterol, homocysteine, fibrinogen, atrial fibrillation, carotid imaging, previous transient ischemic attack or ischemic stroke, peripheral arterial disease, coronary artery disease, congestive heart failure and smoking. Through the step-by-step analysis, we deleted the risk factors without predictive value for recurrent stroke of patients in derivation group, and we introduced the risk factors which predict recurrent stroke of patients in derivation group. And we determined the best weighting scheme of each independent risk factor to make the scoring model get the maximum C-statistic. We finally developed a SRS scoring model, seeing table (1).(Ⅱ) Assessment of predictability:In the derivation group, for stroke recurrence risks at30days,90days,180days,1year,2years, ESRS predicted the risk of recurrent stroke well (C-statistics0.6063～0.6936). The ABCD2score predicted poor (C-statistics0.3970～0.5702), SRS (C-statistics0.5863～0.6940) was similar with ESRS, both SRS and ESRS were better than ABCD2score. In the validation group, for stroke recurrence risks at30days,90days,180days,1year, the ABCD2score modestly predicted the risk of recurrent stroke (C-statistics0.5329～0.5790), SRS and ESRS similarity predicted the risk of recurrent stroke (C-statistics0.4726～0.5335,0.4656～0.5019, respectively). In the pooled group, for stroke recurrence risks at30days,90days,180days,1year, SRS and ESRS similarily predicted the risk of recurrent stroke (C-statistics0.5438～0.5925,0.5487～0.5884, respectively), both SRS and ESRS were better than the ABCD2score (C-statistics0.4957～0.5728). In the TIA group, for stroke recurrence risks at30days,90days,180days,1year, SRS was similar with the ABCD2score (C-statistics0.3047～0.5000,0.4703～0.5391, respecttively), both SRS and the ABCD2score were better than ESRS (C-statistics0.1719～0.4209). In the risk assessment of stroke recurrence in the pooled group, for the accuracy of the scoring models assessed high-risk patients as risk groups at30days,90days,180days,1year, the sensitivity and specificity of SRS were77.8～82.7%and33.2～35.2%, respectively. ESRS:66.7～73.0%and40.1～41.7%. ABCD2:89.2～100%and8.8～9.3%. Except the validation group and the TIA group, for stroke risk assessment in the derivation group, and the pooled group, patients, which were assessed with high risk by SRS and ESRS, had high stroke recurrence rate (P<0.05), but SRS had better performance than ESRS. Patients, which were assessed with low-risk or high-risk by the ABCD2score, had similar stroke recurrence risks (P>0.05).Conclusions:In the groups of acute ischemic cerebrovascular disease mainly with acute ischemic stroke patients, SRS and ESRS had similar predictive accuracy for stroke recurrence risk, but the overall prediction accuracy of SRS was slightly higher than ESRS. And not surprisingly, the ABCD2score had much poor prediction accuracy than both SRS and ESRS. Although SRS and ESRS had similar differentiation ability for patients at high risk of stroke recurrence, the overall differentiation ability of SRS was slightly higher than ESRS. The ABCD2score had poor differentiation ability. The results and conclusions should be further confirmed in prospective studies.