OBJECTIVE To compare three-dimensional (3D) time-of-flight (TOF) MR angiography, contrast-enhanced MR angiography, digital subtraction angiography (DSA), and rotational angiography for depiction of stenosis. METHODS The study had Ethics Committee approval, and each patient gave written informed consent. Forty-nine patients (18 women, mean age, 67.2 years +/- 9.1 [+/- standard deviation], and 31 men, mean age, 63.1 years +/- 8.0) with symptomatic stenosis of internal carotid artery (ICA) diagnosed at duplex ultrasonography underwent transverse 3D TOF MR angiography with sliding interleaved kY acquisition and coronal contrast-enhanced MR angiography, followed by DSA and rotational angiography within 48 hours. MR angiography was performed at 1.5-T with a cervical coil. Contrast-enhanced MR angiograms were obtained after a bolus injection of 20 mL of gadobenate dimeglumine. Maximum ICA stenosis on maximum intensity projection and source images was quantified according to NASCET criteria. Correlations for 3D TOF MR angiography, contrast-enhanced MR angiography, DSA, and rotational angiography were determined by means of cross tabulation, and accuracy for detection and grading of stenoses were calculated. Data were evaluated with analysis of variance, Wilcoxon signed rank test, and McNemar test, all at significance of P < .05. RESULTS Ninety-eight ICAs were evaluated at contrast-enhanced MR angiography, DSA, and rotational angiography, and 97 were evaluated at 3D TOF MR angiography. Correlations for contrast-enhanced MR angiography, 3D TOF MR angiography, and DSA relative to rotational angiography were r2 = 0.9332, r2 = 0.9048, and r2 = 0.9255, respectively. Lower correlation (r2 = 0.8593) was noted for contrast-enhanced MR angiography and DSA. Respective sensitivity and specificity for detection of hemodynamically relevant stenosis relative to rotational angiography were 100% and 90% for contrast-enhanced MR angiography, 95.5% and 87.2% for 3D TOF MR angiography, and 88.6% and 100% for DSA. Four of 31 severe stenoses were underestimated at DSA, and three were underestimated at contrast-enhanced MR angiography. Three severe stenoses were underestimated at 3D TOF MR angiography, and one was misclassified as occluded. Of 13 moderate (50%-69%) stenoses, one was overestimated at contrast-enhanced MR angiography, two were underestimated and three overestimated at 3D TOF MR angiography, and two were underestimated at DSA. CONCLUSIONS DSA results in an underestimation of ICA stenosis compared with rotational angiography. Contrast-enhanced MR angiography correlates best with rotational angiography.