Blinded-reader comparison of magnetic resonance angiography and duplex ultrasonography for carotid artery bifurcation stenosis

We compared two-dimensional time-of-flight magnetic resonance angiography (MRA) and duplex ultrasonography with arteriography for the detection of 70% to 99% stenoses at the carotid artery bifurcation (ie, surgical disease according to findings of the North American Carotid Endarterectomy Trial). Three blinded readers independently measured stenoses on MRA in 73 vessels from 38 patients. Duplex ultrasonography was available in 66 vessels from 35 of these patients, and blinded reading was performed by one reader. Comparison was made to arteriography. Magnetic resonance angiography demonstrated a sensitivity of 92.4%, specificity of 74.5%, and negative predictive value of 95.8% for 70% to 99% stenoses. Interobserver agreement was high (kappa = 0.91). Absence of signal at stenoses with evidence of distal flow usually, but not always, corresponded to surgical disease. Duplex ultrasonography demonstrated a sensitivity of 81.0%, specificity of 82.2%, and negative predictive value of 90.2% for surgical disease. There was no significant difference between MRA and duplex ultrasonography for the sensitivity or specificity in detecting 70% to 99% stenoses (P > .1, exact form of the McNemar test). MRA had no false positives or false negatives for complete occlusions of the carotid artery, whereas duplex ultrasonography missed one occlusion and falsely called two patent vessels occluded. In seven cases, both MRA and duplex ultrasonography overestimated stenoses to miscategorize them as surgical disease. Although the sensitivity and specificity of MRA and duplex ultrasonography are not significantly different for distinguishing surgical and nonsurgical degrees of stenosis at the carotid bifurcation, MRA has some advantages that may make it the screening test of choice. Concordant MRA and duplex ultrasonography for surgical disease does not necessarily obviate the need for catheter arteriography.