Purpose: The purpose of this study was to evaluate the carotid duplex criteria for a > or = 60% angiographic internal carotid artery (ICA) stenosis and the degree of variation among duplex scanners.
Methods: Carotid duplex criteria for a > or = 60% angiographic stenosis were evaluated in two ICAVL-accredited vascular laboratories with different brands of duplex scanners (Siemens-Quantum and Diasonics in Laboratory A, ATL and Diasonics in Laboratory B). Analysis was performed for 360 carotid bifurcations in 180 consecutive patients who had concurrent angiographic and duplex evaluation. Blinded angiogram evaluation was performed with precision electronic calipers on magnified views, with stenosis calculated by criteria of the Asymptomatic Carotid Atherosclerosis Study and the North American Symptomatic Carotid Endarterectomy Trial. Duplex data included internal carotid artery peak systolic velocity (ICA PSV), ICA end-diastolic velocity, and the ratio of ICA PSV to common carotid artery (CCA) PSV (ICA/CCA ratio).
Results: The most accurate determination of a > or = 60% ICA stenosis was obtained with ICA/CCA ratio and ICA PSV, but the optimal threshold differed for all four scanners. The optimal ICA/CCA ratio varied from 2.6 to 3.3, and the optimal ICA PSV varied from 190 to 240 cm/sec. All four scanners produced criteria that give a positive predictive value > 90% while maintaining accuracy at > or = 90%. Logarithmic transformation of duplex variables created a linear relationship between duplex values and angiographic stenosis, allowing statistical evaluation of scanner operating characteristics by linear regression analysis and analysis of covariance. This analysis revealed that the mathematic equation relating duplex values with angiographic percent stenosis was statistically different for one of the four scanners (p < 0.05). Scanner differences did not appear to be due to technologists, because the regression lines were nearly identical for the two Diasonics scanners despite use by different technologists. Ignoring the significant difference in operating characteristics for one of the four scanners would result in a mean error for predicting a 60% stenosis of 14% to 18% (equating a 46% or 78% stenosis with a 60% stenosis).
Conclusions: We conclude that the correlation of duplex data with angiographic percent stenosis and the duplex criteria for a > or = 60% stenosis are machine-specific. Regression analysis can determine whether apparent differences are due to chance or significant differences in scanner characteristics. Future studies should include regression analysis according to equipment type.