Related Experiment Video
Updated: Jul 10, 2026
Combined Near-infrared Fluorescent Imaging and Micro-computed Tomography for Directly Visualizing Cerebral Thromboemboli
Published on: September 25, 2016
High-specificity identification of large vessel occlusion stroke using D-dimer and NT-proBNP combined with clinical
Leire Azurmendi1, J Gonzalez-Fraile2, A Barragán-Prieto3
1Department of Internal Medicine, Faculty of Medicine, Geneva University, Genève, Switzerland.
Background:
Rapid identification of large vessel occlusion (LVO) is essential to optimize prehospital triage and timely access to mechanical thrombectomy, yet current clinical scales show limited diagnostic performance, particularly when high specificity is required to ensure patient safety.
Methods:
We performed a prospective analysis of 290 consecutive patients with suspected acute ischemic stroke presenting within 24 h of symptom onset (n = 96 LVO), including 223 patients presenting within 6 h (n = 71 LVO). Presence of LVO was determined by vascular neuroimaging (computed tomography angiography or magnetic resonance angiography) performed during routine clinical care. Circulating blood biomarkers and clinical variables were systematically evaluated to derive a multivariable panel for LVO detection. A data-driven modeling strategy (PanelomiX) was applied under predefined high-specificity constraints, with repeated stratified cross-validation used for panel ranking and robustness assessment. Diagnostic performance was compared against individual variables and previously reported biomarker-clinical combinations.
Results:
A multimodal panel combining NT-proBNP, D-dimer, mean blood pressure, and the National Institutes of Health Stroke Scale (NIHSS) score demonstrated superior discrimination compared with individual biomarkers, clinical variables alone or any other considered biomarker-clinical combination. At a specificity of 90%, the panel achieved a sensitivity of 59.8% in the 24 h cohort. Among patients presenting within 6 h of symptom onset, sensitivity reached 65.7% while maintaining a specificity of 90.3%. Cross-validated Youden index estimates decreased to 0.44 in both cohorts, confirming modest optimism bias consistent with the exploratory design of this single-cohort study.
Conclusion:
The integration of circulating biomarkers with clinical variables improved the high-specificity identification of LVO. This multimodal approach should be interpreted as a rule-in enrichment strategy to support early triage and optimized routing of patients with a high probability of LVO, rather than as a rule-out tool to exclude LVO in lower-risk patients.