Accurate patient selection based on preoperative imaging is imperative to good risk reduction in patients undergoing carotid endarterectomy (CEA). The goal of this study was to assess the accuracy of gadolinium-enhanced magnetic resonance angiography (GE MRA) versus time-of-flight (TOF) MRA in the work-up of patients undergoing CEA. Patients undergoing CEA between 1999 and 2001 were identified from a prospectively maintained institutional database. GE or TOF MRA was obtained on extracranial carotid arteries (n = 319) in patients undergoing CEA. Stenosis on MRA images was graded as moderate (n = 76) or severe (n = 243) by an attending radiologist who was blind to duplex results. Duplex imaging was performed in an Intersocietal Commission for the Accreditation of Vascular Labs (ICAVL) accredited lab, and stenosis was stratified as moderate (50-79%, n = 76) or high (80-99%, n = 243) grade using University of Washington criteria. For each patient, the degree of stenosis as determined by MRA (GE versus TOF) was compared to percent stenosis on duplex. For moderate-grade lesions, GE MRA concurred with duplex in 11.1% (4/36), underestimated in 2.8% (1/36), and overestimated in 86.1% (31/36) of carotid arteries imaged. TOF MRA concurred with duplex in 35% (14/40), underestimated in 0% (0/40), and overestimated in 65% (26/40) of carotid arteries. High-grade lesions demonstrated improved concordance between MRA and duplex. For these lesions, GE MRA concurred with duplex in 95.6% (130/136) of carotid arteries imaged, never overestimated stenosis (0/136), and underestimated in 4.4% (6/136). TOF MRA concurred with duplex 96.3% (103/107), overestimated stenosis as an occlusion in 0.9% (1/107), and underestimated in 2.8% (3/107). In addition to neck visualization, the GE technique allowed simultaneous aortic arch imaging. This was accomplished in 79.1% (136/172) of all GE MRAs. Simultaneous aortic arch imaging was not technically feasible with TOF MRA. For moderate-grade lesions, both MR techniques are inaccurate predictors of degree of carotid stenosis and result in a significant overestimation of stenosis. Each technique demonstrates improved concordance with duplex ultrasound in the setting of severe carotid artery stenoses. The ability of GE MRA to simultaneously image the aortic arch and the neck may allow for detection of occult tandem lesions and other anatomic variations, which may be particularly important in preoperative planning for carotid artery stenting.