Abstract
INTRODUCTION: Stroke remains a leading cause of death and disability worldwide, and despite advances in reperfusion therapies, early and accurate prediction of stroke severity and outcomes remains suboptimal. We examine the associations between eosinophil levels and stroke severity, functional outcomes, mortality, post‐stroke infections, and hemorrhagic transformation. METHODS: We conducted a systematic review in accordance with PRISMA. Electronic searches of PubMed, Embase, Web of Science, and Google Scholar through February 2025, including adult cohort studies of acute stroke that reported admission eosinophil counts or percentages alongside odds ratios for poor functional outcome (mRS ≥3), mortality, symptomatic intracerebral hemorrhage, or post‐stroke infection. Statistical analyses were performed using the R software with random‐effects meta‐analysis of log‐transformed ORs, heterogeneity was assessed via Cochran's Q and I(2), and subgroup analyses were performed by stroke subtype. RESULTS: Twenty‐one cohort studies (n = 22 930) met inclusion criteria, with most focused on ischemic stroke and varied eosinopenia cut‐off (0.005‐0.045 × 10⁹/L). Across all stroke types, eosinopenia was not significantly associated with poor 90‐day functional outcome (pooled OR 1.80; 95 % CI 0.82‐3.92; I(2) = 90.7 %), but in ischemic stroke cohorts alone, eosinopenia doubled the odds of poor outcome (OR 2.32; 95 % CI 1.06‐5.11; I(2) = 91.1 %). Lower eosinophil levels were associated with a 2.75‐fold increase in mortality risk (95 % CI 1.44‐5.27; I(2) = 75.9 %) and symptomatic intracerebral hemorrhage (OR 4.53; 95 % CI 2.46‐8.36; I(2) = 0 %). No statistically significant association was found with post‐stroke infection (OR 1.40; 95 % CI 0.49‐3.97; I(2) = 83.5 %). CONCLUSION: Admission eosinophil level shows promise as an adjunctive prognostic biomarker particularly for ischemic stroke outcomes, early mortality, and hemorrhagic transformation but its moderate predictive strength and inconsistent infection associations warrant its use alongside established models. Large, standardized prospective studies and mechanistic investigations are needed to validate optimal thresholds, timing, and integration into multi‐marker prognostic algorithms before routine clinical implementation.