In recent years, recurrent ischemic stroke has become the focus of growing scientific research and clinical expertise. Modern approaches increasingly use artificial intelligence (AI) systems in conjunction with sophisticated mathematical and statistical algorithms, especially machine learning models, to create individualized patient profiles and produce customized predictions about the likelihood of subsequent cerebrovascular events.^[1]

Kolmos et al. identified the most significant independent risk factors for recurrent cerebral infarction, including arterial hypertension, atrial fibrillation and other arrhythmias and conduction disturbances, diabetes mellitus, smoking, prior cerebrovascular events, and complicated clinical presentations during the initial stroke episode.^[2]

Current evidence on sex-related differences in recurrent cerebrovascular risk remains inconsistent. The POINT (Platelet-Oriented Inhibition in New TIA and Minor Ischemic Stroke) randomized controlled trial, which enrolled 4,881 patients with minor ischemic stroke or high-risk transient ischemic attack (TIA) and followed them for 90 days, found no statistically significant difference in risk between women (5.76%) and men (5.67%). Although a trend toward lower recurrence risk was observed among women under the age of 60, the difference did not reach statistical significance.^[3]

According to Fu et al., the most common risk factors included age, male sex, a history of cardiovascular or cerebrovascular events, diabetes mellitus, and coronary artery disease.^[4] In a retrospective study involving 787 patients conducted at a single hospital center, Chung et al. identified arterial hypertension and dyslipidemia as the primary predictors of subsequent acute stroke, independent of sex. However, sex-specific associations were observed: diabetes mellitus was a prominent metabolic risk factor in women, while smoking and alcohol consumption were major behavioral risk factors in men.^[5]

The objective of the study was to make a comparative retrospective analysis of non-modifiable and modifiable risk factors and clinical characteristics in elderly patients with a first-ever, second, third and subsequent ischemic stroke.

This study is part of previous original research.^{[6, 7]} A total of 260 patients (aged >65 years) diagnosed with acute stroke and admitted to the Neurology Clinic between January 1, 2023, and March 1, 2024, were included. For the purposes of the analysis, patients were divided into three groups: Group 1—first-ever stroke (n=212), Group 2—second stroke (with a history of prior cerebrovascular event) (n=35), and Group 3—third and recurrent stroke (with a history of at least two previous cerebral infarctions) (n=13). A model incorporated an assessment of the most prevalent vascular risk factors: arterial hypertension, diabetes mellitus, dyslipidemia, cardiovascular disease, and smoking

Two etiological classification systems were applied: the Trial of Org 10172 in Acute Stroke Treatment (TOAST)^[8] and the ASCOD phenotyping system^{[9, 10]}. Stroke severity was assessed using the NIH Stroke Scale (NIHSS)^[11], and cognitive impairment was evaluated using the Mini-Mental State Examination (MMSE)^[12].

The TOAST classification identified five primary etiological subtypes: 1. Large-artery atherosclerosis (involving the aorta, carotid, and vertebral arteries), 2. Cardioembolism, 3. Small-vessel diseases, 4. Other determined causes (e.g., migraine, vasculitis, vasculopathies, and venous thrombosis), and 5. Stroke of undetermined etiology (cryptogenic).^[8]

According to the ASCOD phenotyping system, patients were divided into five etiological profiles: arterial dissection (D), cardioembolic source (C), small-vessel disease (S), atherosclerosis (A), and other causes (O). Each category was further graded according to defined diagnostic criteria and levels of causal likelihood, corresponding to three levels of evidence: A (definite), B (probable), and C (possible). The degree of phenotypic expression was classified as follows: 0 – absence of disease, 1 – presence of a potential factor, 2 – presence of a possible but uncertain factor, 3 – presence of a non-etiological but comorbid factor, and 9 – insufficient clinical and instrumental data to allow interpretation.^{[9, 10]}

NIHSS scores were grouped into three severity categories: mild (0–7), moderate (8–15), and severe (>16).^[11] The affected vascular territories were defined as the left middle cerebral artery (MCA), right MCA, and vertebrobasilar system. Cognitive impairment was determined by an MMSE score ≤ 27 points.

Statistical analyses included descriptive statistics, recoding of variables, independent-samples t-tests, chi-square tests, one-way analysis of variance (ANOVA), multiple linear regression, multinomial logistic regression, decision tree modeling using Classification and Regression Trees, radial basis function neural networks, and graphical data representations.^{[13, 14]} A significance level of 0.05 (p-value) was adopted. Data were analyzed using IBM SPSS Statistics, version 26.

In the studied cohort, the incidence of recurrent cerebrovascular events was 18.46%. Analysis of the age distribution across the three groups revealed a clear progressive trend with the average age increasing from Group 1 to Group 3. When stratified by sex, females in each group exhibited a higher average age compared to males. The regression and classification coefficients confirmed age as an independent vascular risk factor with strong predictive value, in line with findings from numerous previous studies.^[15–18]

Significant independent comorbidities exerted a moderate influence on the risk of recurrent stroke, including dyslipidemia, cardiac rhythm and conduction abnormalities, other cardiovascular diseases and diabetes mellitus, corroborating the results of Callaly et al. and Elhefnawy et al.^{[19, 20]} Multinomial logistic regression analysis comparing Groups 1 and 2 identified dyslipidemia, in addition to age, as a statistically significant factor associated with an increased risk of recurrent stroke.

No statistically significant differences were found among the three groups with respect to sex, arterial hypertension, smoking status, and affected vascular territories. Clinically, patients with second and subsequent strokes demonstrated more severe residual motor and/or speech impairments and cognitive deficits, correlating with poorer prognostic outcomes.^[21]

In a study by Elnady et al., based on the TOAST classification, the cumulative risk of recurrent stroke was found to be highest and occurred earliest (<1 year) in cases of large-artery atherosclerosis and cardioembolism, whereas it was lowest and occurred later (>1 year) in cases attributed to small-vessel disease and cryptogenic stroke.^[22] Uzuner and Uzuner analyzed four primary TOAST subtypes in the context of vascular recurrences: cardioembolism, large-artery atherosclerosis, strokes of undetermined etiology, and small-vessel disease.^[23] Our results indicated that large-artery atherosclerosis was the primary etiology in Groups 2 and 3, followed by cardioembolism, small-vessel disease, and cryptogenic stroke.

The ASCOD classification eliminated the term ‘cryptogenic’ and instead evaluated coexisting conditions in cases of possible Embolic Stroke of Undetermined Source (ESUS). It has demonstrated strong predictive value in estimating the risk of subsequent stroke.^{[9, 10, 24]} Among patients in Groups 2 and 3, phenotypes A, S, and C were identified as significant etiological factors. Categories A1, A2, S3, C1, and C2 correlated with an increased incidence of second, third, and recurrent cerebral infarctions.

The structure of the artificial neural network supports the hypothesis that the inclusion of sex as a variable, along with elements from the ASCOD system, may enhance the accuracy and effectiveness of predictive models.

This study is subject to several limitations, notably the presentation of preliminary data, a short follow-up period, a small number of cases in Groups 2 and 3, and the lack of a comparative analysis of secondary prevention strategies.

Recurrent ischemic stroke is associated with a moderate to high prevalence. Rigorous quantification of both cumulative and individual vascular risk is essential for guiding therapeutic strategies, optimizing secondary prevention efforts, and establishing a robust foundation for long-term prognosis in cerebrovascular disease progression.

The authors have no funding to report.

The authors have declared that no competing interests exist.

All authors have approved the manuscript and agree with its submission to your journal.