Purpose: The North American Symptomatic Carotid Endarterectomy Trial (NASCET) and the Asymptomatic Carotid Artery Study (ACAS) both confirmed the effectiveness of carotid endarterectomy for preventing stroke in patients who have significant carotid stenosis. A uniform technique for measuring carotid stenosis from an arteriogram (% stenosis = [1 - minimum residual lumen/normal distal cervical internal carotid artery diameter] x 100) was used in both trials, with reproducibility internally validated. The reliability of this measurement when used outside the trials for defining carotid stenosis has not been validated. Imprecise calculation of carotid stenosis can result in a 50% overestimation of significant carotid disease and potential overuse of carotid surgery. This is a prospective study of the reliability of carotid stenosis measurements performed by practicing physicians of different specialties and different levels of clinical experience.
Methods: Two vascular surgeons and two interventional radiologists (one resident and one staff member per specialty), blinded to results, calculated the percent stenosis from 219 consecutive arteriograms performed to evaluate extracranial carotid artery occlusive disease; 72 random films were reread by each individual. The interpretations were grouped as < 60% or > or = 60% stenosis (ACAS) and as < 30%, 30% to 69%, and > or = 70% stenosis (NASCET). Interobserver and intraobserver agreement were analyzed with the kappa statistic and Pearson correlation coefficients.
Results: Interobserver reliability in categorizing carotid stenosis revealed excellent agreement for both ACAS (kappa = 0.825 to 0.903) and NASCET groups (kappa = 0.729 to 0.793). Interobserver correlation coefficients ranged from 0.91 to 0.95. Intraobserver agreement was also highly reproducible for both the ACAS (kappa = 0.732 to 0.970) and NASCET categories (kappa = 0.634 to 0.805). Intraobserver correlation coefficients ranged from 0.89 to 0.95.
Conclusion: The NASCET technique for quantification of carotid stenosis can be easily learned by physicians and reliably implemented for appropriate identification of candidates for carotid endarterectomy.