Clinical carotid endarterectomy decision making: noninvasive vascular imaging versus angiography.
OBJECTIVE: Carotid endarterectomy (CEA) is frequently performed based solely on noninvasive vascular imaging (NVI) results (duplex ultrasound, DU; magnetic resonance angiography, MRA; CT angiography, CTA). The authors determined how often intra-arterial contrast angiography (ANGIO) alters a CEA decision as compared to NVI in clinical practice. METHODS: Reports of all NVI studies in 569 consecutive patients undergoing ANGIO at an academic medical center (AMC, n = 360) and a community hospital (CH, n = 209) over 3 years were reviewed. Patients were classified as to whether CEA was indicated based on each study. Misclassification rates, sensitivities, specificities, positive (PPV) and negative (NPV) predictive values were calculated. RESULTS: CTA was performed infrequently (2.5%) and not considered further. Misclassification rates for CEA based on DU in the AMC and CH were similar. The misclassification rate for DU alone was 28% (95% CI: 24,32), and for MRA alone was 18% (95% CI: 11,25). Both NVI were done in 11% of patients, with a misclassification rate of 7.9% (95% CI: 0,16) when the two were concordant (76% of studies). DU had a sensitivity of 87% (95% CI: 83,91), specificity 46% (95% CI: 38,54), PPV 73% (95% CI: 68,78) and NPV 68% (95% CI: 60,77). MRA had a sensitivity of 75% (95% CI: 63,87), specificity 88% (95% CI: 80,96), PPV 84% (95% CI: 73,95) and NPV 80% (95% CI: 70, 90). The sensitivity of concordant NVIs was 96% (95% CI: 88,100), specificity 85% (95% CI: 65,100), PPV 93% (95% CI: 81,100) and NPV 92% (95% CI: 76,100). CONCLUSION: These data suggest that surgical decisions should be made with caution if based on the results of noninvasive studies, particularly DU performed alone. Concordant DU and MRA results in a lower misclassification rate than either test used alone.
Johnston, DC; Goldstein, LB
Volume / Issue
Start / End Page
Pubmed Central ID
International Standard Serial Number (ISSN)
Digital Object Identifier (DOI)