Background: Identifying large vessel occlusion (LVO) is paramount in the era of endovascular therapy. This can be achieved easily by computed tomography angiogram(CTA) and magnetic resonance angiography(MRA). National Institute Health Stroke Scale(NIHSS) threshold can be used in primary stroke centre(PSC) without these facilities, to identify patients with LVO for transfer to a comprehensive stroke centre (CSC) for mechanical thrombectomy. Alternatively, simplified clinical stroke scales (CSS)can also be used by emergency medical services (EMS) to identify such patients to be directly transported to CSC. We aim to compare various CSS in predicting LVO in ischemic stroke.
Methods: We searched PUBMED from January 2005 to July 2016 and screened reference lists of included studies, and included all diagnostic accuracy studies that investigate CSS and prediction of LVO confirmed with CTA or MRA. Two authors independently screened titles and abstracts and perform data extraction for analysis. Pooled sensitivity, specificity and diagnostic odd ratio (DOR) of various CSS were obtained using the random effects model.
Result: 8 studies (total 8556 patients) were included for meta-analysis of 3 CSS. This includes the NIHSS≥5, ≥6, ≥10, ≥13; Cincinnati Prehospital Stroke Scale (CPSS)≥2 and Rapid Arterial oCclusion Evaluation (RACE) ≥ 5. Pooled sensivity and specificty for the respective cut off values of these CSS was metaanalysed and presented in table 1. NIHSS≥13 has the highest DOR (26.0), followed by NIHSS≥5 (16.9); whereas other NIHSS thresholds, CPSS and RACE were less predictive. NIHSS≥5 was found to have the highest sensitivity of 0.865 and NIHSS≥13 had the highest specificity of 0.934.
Conclusion: We found 2 NIHSS thresholds that were most useful for identifying LVO. Depending on resources and preferences of each stroke centre, different NIHSS threshold can be utilized for purpose of ruling in LVO using a high NIHSS threshold, or ruling out LVO using a low NIHSS threshold.