Development of a Clinical Prediction Model for Ultra-Early Mild Acute Ischemic Stroke: A Novel Approach to Early Diagnosis and Treatment
Introduction:
Cerebrovascular disease, particularly acute ischemic stroke (AIS) caused by cerebral atherosclerosis, remains a significant health concern in China. AIS, characterized by localized blood flow interruption to the brain, leads to various neurological impairments. AIS is the most prevalent stroke subtype and is associated with high disability and mortality rates.1,2 Intravenous thrombolysis within 6 hours of symptom onset has shown to improve neurological outcomes in AIS patients.3-5 However, the narrow therapeutic window necessitates rapid and accurate diagnosis.6 Transient ischemic attack (TIA), often a precursor to stroke, shares similar pathological mechanisms and is managed with antiplatelet agents, antithrombotic therapy, and cerebral perfusion enhancement.7
Current clinical guidelines acknowledge the challenge in distinguishing CT-negative ultra-early mild AIS from TIA based on clinical presentation alone. MRI with diffusion-weighted imaging (MRI-DWI) is the gold standard for differentiation but is often inaccessible in primary hospitals due to high costs and time constraints.8 Consequently, CT is widely used for initial assessment. However, CT has limited sensitivity in detecting early ischemic changes, with high false-negative rates in mild AIS cases, potentially delaying thrombolytic therapy.9
Biomarkers and Their Role:
In recent years, serum biomarkers have gained attention for their roles in ischemic stroke pathogenesis and prognosis. Inflammation, endothelial dysfunction, and metabolic alterations play critical roles in acute cerebral ischemia.10 Markers such as high-sensitivity C-reactive protein (hs-CRP), homocysteine (HCY), and lipid profiles have been associated with stroke risk and outcomes.11,12 Moreover, dynamic changes in neutrophil-to-lymphocyte ratio (NLR) and platelet-to-lymphocyte ratio (PLR) may reflect the acute inflammatory state following cerebral ischemia.13
Model Development and Validation:
This study aimed to create and validate a clinical prediction model integrating NIHSS scores with readily available serum biomarkers to distinguish CT-negative mild AIS from TIA at an early stage. The goal is to offer a practical and quick diagnostic tool for clinical decision-making in MRI-unavailable settings. The model aims to enhance timely interventions and improve patient outcomes.
Materials and Methods:
The study included patients with CT-negative ultra-early mild AIS and TIA admitted to a comprehensive hospital in Shishi City, China, between 2020 and 2023. Mild AIS was defined by mild neurological deficits with specific scoring thresholds (NIHSS ≤ 5). TIA is typically defined as a temporary blood flow disruption causing symptoms lasting <24 hours with minimal lasting neurological damage. Symptoms are generally mild and may include brief weakness, numbness, speech difficulties, or visual disturbances. Unlike stroke, TIA resolves completely. The hospital's high-quality medical services and well-organized resources made it a representative institution for the area.
Inclusion and Exclusion Criteria:
Inclusion criteria included suspected AIS or TIA patients within 6 hours of onset, CT examination without responsible ischemic lesions, confirmed AIS or TIA by MRI-DWI, NIHSS scores ≤ 5, and peripheral blood collection within 2 hours of admission for analysis and tests. Exclusion criteria covered incomplete information, infectious diseases history, stroke or TIA history, craniocerebral trauma, craniocerebral surgery history, hematologic diseases, malignancies, inability to cooperate with MRI, autoimmune diseases, HIV or immunosuppressants, arrhythmias, and incomplete information.
Sample Size Estimation:
The study's sample size was determined using the empirical rule for multivariate regression models, considering the number of variables. This required a sample size 10-20 times the number of independent variables. After literature review, theoretical analysis, and pre-test screening, 6 independent variables were finalized. The estimated sample size range was 60 to 120 cases, and the planned sample size was set at 330 cases, meeting statistical requirements for model development.
Laboratory Measurements:
Complete blood count (CBC), C-reactive protein (CRP), D-dimer, lipid profiles, and glucose levels were measured using specific analyzers and reagents. Variables included gender, age, hypertension history, diabetes history, NIHSS score, HCY, neutrophil count, lymphocyte count, monocyte count, platelet count, CRP, fibrinogen, D-dimer, GLU, TG, HDL, LDL, NLR, and PLR.
Statistical Analysis:
Data were randomly split into training and validation sets. Statistical analyses were performed using SPSS Statistics software. Normality was assessed using the Kolmogorov-Smirnov test. Normally distributed data were compared using t-tests, while non-normal data used Mann-Whitney U-tests. Categorical data were compared using chi-square tests.
Results:
Baseline characteristics were compared between AIS and TIA subgroups. Multivariate logistic regression identified independent risk factors. A clinical prediction model was developed using R software, evaluated for discrimination, calibration, and clinical utility. A nomogram was constructed for visualization.
Discussion:
The model demonstrated strong discriminative ability, good calibration, and clinical utility across training and validation sets. Multivariate analysis revealed NIHSS, CRP, glucose, total cholesterol, triglycerides, and LDL as independent predictors of mild AIS. The NIHSS score emerged as the strongest predictor, emphasizing its role in quantifying early neurological deficit and prognostic value in stroke outcomes.
Limitations and Future Directions:
The study had limitations, including a single-center design, small sample size, and lack of emerging biomarkers like GFAP or S100B. Future studies should validate the model in multi-center, prospective cohorts and assess its impact on clinical decision-making and healthcare resource utilization.
Conclusion:
This prediction model offers a practical, evidence-based tool for identifying CT-negative ultra-early mild AIS in resource-limited settings. Its integration into clinical workflows may accelerate thrombolysis initiation, reduce diagnostic delays, and improve functional recovery.
