RESUMEN
PURPOSE: To evaluate the accuracy of multidetector-row CT angiography (CTA) for the diagnosis of large-vessel vasospasm following subarachnoid hemorrhage by comparison to digital subtraction angiography (DSA). METHODS: Thirty-three patients with acute subarachnoid hemorrhage were enrolled in a prospective study and underwent a total of 40 CTA and DSA examinations within 24 h of each other. Two neuroradiologists reviewed the CTA examinations independently. A third neuroradiologist blinded to the CTA results reviewed the DSA examinations. In each patient, for both techniques, 23 arterial segments were evaluated for their degree of narrowing; the reviewers were asked to attribute every narrowing to 'vasospasm' or 'hypoplasia'. Agreement between CTA and DSA for the degree of narrowing, and agreement between the two CTA readers, were calculated using weighted kappa-coefficients. Sensitivity, specificity, accuracy, positive and negative predictive value (NPV) of CTA to detect large-vessel vasospasm were calculated considering DSA as the gold standard. RESULTS: Substantial correlation (kappa = 0.638) was found between CTA and DSA for the detection of arterial narrowing. Interobserver agreement between the two CTA reviewers for the degree of luminal narrowing was substantial (kappa = 0.712).CTA was 87% accurate for the diagnosis of large-vessel vasospasm; the NPV of CTA was 95%. CTA was more accurate, and interobserver agreement higher, for the proximal arterial segments (basilar and vertebral arteries) than for the distal ones (P2 segments). Using CTA as a screening modality, 83% of unnecessary DSA would have been avoided. CONCLUSION: Compared to the gold standard of DSA, CTA is accurate for the detection of large-vessel vasospasm, and has a very high NPV.