Background: Benefit from carotid endarterectomy (CEA) centers on patient selection and percent stenosis as determined by cerebral angiography. However, angiography remains expensive and poses risks. Validated carotid duplex ultrasonography has proven to be an accurate tool for selecting patients for CEA. However, the role of another noninvasive test-magnetic resonance angiography (MRA)-remains uncertain. Because of recent advances in MRA hardware and software, we hypothesized that clinically appropriate patients could be accurately selected for CEA based on MRA alone.
Methods: Fifty-four carotid arteries in 29 patients (with and without symptoms) underwent both three-dimensional time-of-flight MRA (1.5 Tesla) with multiple overlapping thin slab acquisition and biplanar intra-arterial digital subtraction angiography. All patients undergoing both tests over a 24-month period were included. The majority of these patients did not undergo carotid duplex ultrasound owing to the clinical practice of the hospital's neurosurgery service. Staff radiologists interpreted each study. The accuracy of patient selection based on MRA was calculated using angiography as the standard (NASCET method). Since operative thresholds vary depending on clinical history, we considered four commonly used ranges of percent stenosis for CEA.
Results: Patient selection accuracy of MRA alone was low, but increased as percent stenosis increased. Out of 10 occluded arteries by angiography, 5 were interpreted as patent with stenosis (70% to 99%) by MRA. One patent artery was misread as occluded on MRA.
Conclusion: Reliance solely on contemporary MRA for surgical decision making cannot be justified in view of low accuracy, which leads to high rates of error in patient selection for CEA.