The Risk of Developing Ischemic Stroke in Patients After Exacerbation of Ischemic Heart Disease

Aim To develop a model for evaluating the risk of stroke in patients after exacerbation of ischemic heart disease who were admitted to the hospitals included into a vascular program. Material This study included 1803 patients with acute coronary syndrome (ACS) from four institutions of Moscow, Kazan, Astrakhan, and Krasnodar where the vascular program was established. Mean age of patients was 64.9±12.78 years, 62,1 % of them were men. The patients were followed up for one year after the discharge from the hospital. External validation of the developed prognostic model was performed on a cohort of patients with ACS included into the RECORD-3 study. Results During the follow-up period, 42 cases of ischemic stroke were observed. The risk of ischemic stroke was associated with the presence of atrial fibrillation (odd ratio (OR) 2.640; р=0.037), diabetes mellitus (OR 2.718; р=0.041), and chronic heart failure (OR 7.049; р=0.011). Protective factors were high-density lipoprotein cholesterol >1 mmol / l (OR 0.629; р=0.041), percutaneous coronary intervention during an index hospitalization (OR 0.412; р=0.042), anticoagulant treatment (OR 0.670; р=0.049), and achieving the blood pressure goal (OR 0.604; р=0.023). The prognostic model developed on the basis of regression analysis showed a good predictive value (area under the ROC curve, 0.780), sensitivity of 80 %, and specificity of 64.6 %. The diagnostic value of other scales for risk assessment was somewhat lower. The area under the ROC curve was 0.692±0.0245 for the GRACE scale and 0.708±0.0334 for CHA 2 DS 2 VASc. In the external validation of the scale based on data of the RECORD-3 study, the diagnostic value was lower although satisfactory as well (area under the ROC curve, 0.651); sensitivity was 78.9 %, and specificity was 52.3 %. Conclusion The study resulted in development of a simple clinical scale, which will probably allow identifying groups at risk of stroke more precisely than with standard scales.

I schemic stroke is an isolated problem associated with severe disability and high mortality among thrombotic complications in patients with a history of acute coronary artery disease (CAD). The rate of ischemic stroke in the GRACE register was approximately 0.34 % within two weeks after the index event [1], and this rate can reach 3 % within a two-year follow-up period [2]. Age, sex, ethnicity, systolic and diastolic blood pressure (BP), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C) and low-density lipoprotein cholesterol (LDL-C), diabetes mellitus (DM), smoking, and antihypertensive, hypolipidemic, and antiplatelet treatment were predictive risk factors (RF) of stroke [3], as well as were factors included in the CHA 2 DS 2 VASc score for assessing the risk of stroke and systemic thromboembolism in patients with atrial fibrillation [4].
However, there are no validated data concerning the risk of stroke in Russian patients with a history of acute CAD. Thus, the objective of this study was to create a model for assessing the stroke risk in patients with a history of acute CAD who were admitted to Russian hospitals with PCI center.

Material and Methods
This study is based on the results of an initiative, openlabel, observational, multicenter study organized by the ORIGINAL ARTICLES § Department of Therapy, Cardiology, and Functional Diagnostics with a course of Nephrology of the Central State Medical Academy of the Administration of the President of the Russian Federation.
For this analysis, we used data from the ORACLE II (ObseRvation after Acute Coronary syndrome for deve-Lop ment of trEatment options; reg. number: clinicaltrials. gov NCT04068909), which included patients with acute coronary syndrome (ACS), who had indications for percutaneous coronary intervention (PCI) due to an index episode. The impossibility of keeping contact with a patient after discharge was the exclusion criteria. All patients signed an informed consent form for inclusion in the study. During 2014-2017, 1,803 patients were included in the study at four hospitals situated in Moscow, Kazan, Astrakhan, and Krasnodar, in which specialized vascular departments were established under the Fight Against Cardiovascular Diseases federal project. The inclusion criteria were described earlier [5]. The Ethics Committee of the Educational and Scientific Medical Center (Central State Medical Academy since 2015) of the Administration of the President of the Russian Federation approved the study (Protocol No. 14 / 14 of October 20, 2014).
The data on adverse outcomes were collected at followup examinations or phone calls made on days 25, 90, 180, and 360 after inclusion in the study. The primary endpoint was all-cause death as verified by autopsy, if any, discharge summaries, outpatient records, and other medical documents. All cases of strokes were analyzed in this part of the study. The type of stroke was verified by medical records and imaging data, if available.
External validation of the resulting prognostic model was performed in a patient cohort of the RECORD-3 trial [6]. The clinical characteristics of the patients are presented in Table 1. The two patient cohorts were comparable in age, sex, and body mass index. Among the patients included in our study, a higher percentage of patients had ST-segment elevation ACS (STE-ACS), a significantly higher percentage of patients had undergone PCI for an index event, and there were more patients with atrial fibrillation (AF) who had received anticoagulants.
Statistical processing of the findings was performed using statistical software suites SPSS 23.0, MedCalc 19.0.3, and STATA 14.0. Nonparametric methods were used for analysis after evaluating the distribution of variables. Means and standard errors (M±SE) were calculated. Discrete values were compared using Pearson's chi-square test. Sensitivity analysis was performed to identify differences between centers in a multicenter trial by using the regression method in the STATA 14.0 statistical software suite. Logistic regression was used to evaluate the independence of the clinical variables. Variables that showed statistical significance in the univariate analysis were included in the multivariate analysis. Results of these analyses are expressed as odds ratio (OR) and 95 % confidence interval (CI). The clinical significance and diagnostic accuracy of the scores tested were analyzed by constructing receiver operator characteristic (ROC) curves for each diagnostic criterion and by calculating the area under curves (AUC). Sensitivity and specificity were calculated for each diagnostic criterion tested. The results were considered significant at p<0.05 for all types of analyses. ORIGINAL ARTICLES §

Results
The follow-up period lasted for 12 mos (median followup was 366 days). During the follow-up period, 228 deaths were registered, including 144 coronary deaths, 14 stroke deaths, 25 decompensated heart failure deaths, and 6 pulmonary embolism deaths. There were 158 repeated episodes of non-fatal ACS. A total of 45 stro kes were reported during the follow -up period in the ORACUL II trial, of which 42 cases were ischemic, and 3 cases were hemorrhagic. Twenty strokes occurred during the initial hospitalization period, ten strokes in the first 9 days after the onset of ACS (including one case of hemorrhagic nature), and the remaining ten episodes from day 10 after discharge from the hospital. Hemorrhagic strokes were reported on days 4, 14, and 50 from the time of ACS. Two patients had AF and received oral anticoagulation  Table 2. Interestingly, patients who had ischemic strokes during the follow-up period were older, more often were hypertensive, and so less often had effectively controlled BP. These patients were more likely to have carbohydrate metabolism disorders, more severe renal dysfunction, peripheral atherosclerosis, AF, and chronic heart failure (CHF). Patients with a history of strokes were less likely to undergo PCI during the index hospitalization and to receive anticoagulant therapy.
Univariate and multivariate logistic regression analyses were performed to assess the impact of various clinical factors on stroke risk. A sensitivity analysis was performed due to possible differences in patient management between the centers. The differences from the reference center (Moscow) were significant for several parameters (Appendix 1). Correction factors were applied during the regression analysis (Appendix 2). The ana lysis results are presented in Table 3.
A clinical model for calculating stroke risk was created using regression analysis. The regression factors were reduced to whole numbers. The calculated scores are presented in Table 4. The resulting model had high diagnos tic significance. The classification was adequate and met the Hosmer -Lemeshow test (χ²=2.465; p=0.789). The area under the ROC curve was 0.780 ( Figure 1). The sensitivity of the score was 80 %, and specificity was 64.6 %.

ORIGINAL ARTICLES §
The stroke risk assessment score was tested in the patient cohort of the RECORD-3 registry (Figure 2). During the 12-mo follow-up period, 33 strokes were reported. The database did not include information about a type of stroke (hemorrhagic or ischemic). Besides, there was no data on the levels of HDL-C. The score's prognostic significance with these limitations was lower with area under the ROC curve of 0.651, the model's sensitivity was 78.9 %, and specificity was 52.3 %. CHA 2 DS 2 VASc and GRACE scores was also used to assess stroke risk. The diagnostic significance of different scores was compared using the area under the ROC curves ( Figure 3). Table 5 shows the parameters of the area under the ROC curve, sensitivity, and specificity of all tested scores.

Discussion
We have analyzed the clinical predictors of stroke after an ACS episode. Based on multivariate regression analysis, our prognostic model included the presence of AF, chronic kidney disease (CKD), DM and CHF, which were RFs, as well as the achievement of target BP levels,     ORIGINAL ARTICLES § high HDL-C levels, PCI during index hospitalization and anticoagulant therapy, which were protective factors.
It should be noted that the majority of these factors have been also associated with stroke risk in the past, including in studies with ACS patients. Thus, according to the American registry, which included more than 73 thousand patients observed in 2008-2011, age, female sex, African-American or Hispanic ancestry, CKD, CHF, DM, and AF were independently associated with the risk of stroke. A history of hyperlipidemia before the episode of ACS was a protective factor [2]. In the international multicenter PROMETHEUS trial, which included 19,914 patients with ACS, 244 strokes were reported during 12 mos of follow-up, i.e., in 1.5 % of patients. A history of stroke was a factor that increased mortality after ACS. A history of cerebrovascular disease, myocardial infarction with or without ST elevation, hypertension, smoking, female sex, and age were factors associated with stroke. Patients with unstable angina had a lower risk of stroke. Statins were prescribed protectively at the time of discharge [7]. Factors such as PCI and anticoagulant therapy did not significantly affect stroke risk in that multivariate analysis, unlike in our study's findings, according to which PCI during index hospitalization was a protective factor. The PROMETHEUS trial analysis showed that patients who had undergone PCI for an episode of ACS were more likely to receive the best possible drug therapy, i.e., antihypertensive and antithrombotic treatment. Mortality and the risk of recurrent cardiovascular complications were lower during the best possible therapy. The number of strokes was also slightly lower. However, the differences were not significant, possibly due to insufficient sample size [8]. Another study showed that patients who had undergone PCI were more compliant with hypolipidemic therapy and more often reached the LDL-C target levels, which may have also contributed to a lower risk of adverse outcomes [9]. Thus, a possible mechanism for PCI's effect on stroke risk is better compliance with drug therapy in patients who have undergone revascularization. The significance of AF as a risk factor of stroke has been repeatedly confirmed in the analysis of two national registers of ACS patients conducted in Italy. There the incidence of stroke was 1.7 % in patients with AF and 0.4 % in patients without AF (p=0.04) [11].
A large meta-analysis that included 14 studies and more than 290 thousand patients showed that newly diagnosed AF in ACS is an independent RF of stroke, especially in patients with STE-ACS. This association remained when groups were standardized by the main RFs included in the CHA 2 DS 2 VASc score [12]. In our study, AF was an independent RF included in the prognostic model. It should be noted that anticoagulant therapy was a protective factor equal to the presence of AF. This may suggest that a significant number of strokes may be of cardioembolic nature in this group of patients.
One of the factors associated with stroke in our prognostic model was HDL-C levels, which were protective. There are no similar data concerning patients with ACS. However, similar trends were previously detected in patients with DM [13], hypertension [14], and in the general population (Multi-Ethnic Study of Atherosclerosis trial) [15].
The resulting clinical score for assessing stroke risk was of high predictive value and had relatively high sensitivity and specificity. It should be noted that the GRACE and CHA 2 DS 2 VASc scores had a slightly lower predictive value in our study. The same scores were tested in a cohort of 4,227 patients with ACS who were followed up for five years after discharge in Spain. During the follow-up period, 183 cases of stroke were reported, and only 16 % of them were associated with AF. The GRACE and CHA 2 DS 2 VASc scores did not have a very high diagnostic value. The areas under the ROC curves were 0.62 and 0.60, respectively. The CHA 2 DS 2 VASc score's sensitivity was higher, but it had lower specificity than the GRACE score [16].

Conclusion
We have developed a relatively simple clinical score that may identify stroke risk groups more accurately than standard scores.

No conflict of interest is reported.
The article was received on 15/05/2020