Determination of 60% or greater carotid stenosis: a prospective comparison of magnetic resonance angiography and duplex ultrasound with conventional angiography.

The morbidity and cost of conventional angiography (CA) have focused recent efforts in cerebrovascular imaging upon the exclusive use of noninvasive techniques. Our purpose was to prospectively evaluate carotid magnetic resonance angiography (MRA) and to compare its accuracy with color-flow duplex (CFD). Fifty patients were prospectively evaluated with CA and MRA after clinical and CFD findings indicated the need for carotid angiography. CFD measurements of peak systolic velocity (PSV) and end-diastolic velocity (EDV) were made. MRA results were categorized as 0%-39%, 40%-59%, 60%-79%, or 80%-99% stenosis or occluded. Determination of percent carotid stenosis by CA was made as in the North American Symptomatic Carotid Endarterectomy Trial (NASCET). Using receiver operating characteristic (ROC) curves, the probability of correctly predicting a > or =60% stenosis using various CFD thresholds and MRA was assessed. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) in determining > or =60% stenosis were estimated. For MRA the sensitivity was 85% (95% Confidence Interval [CI] = 69%-94%), specificity 70% (CI = 56%-81 %), PPV 68% (CI = 53%-80%), and NPV 86% (CI = 72%-94%). For CFD the sensitivity was 89% (CI = 74%-96%), specificity 93% (CI = 82%-98%), PPV 89% (CI = 74%-96%), and NPV 93% (CI = 82%-98%). When MRA and CFD results were concordant (n = 64), the sensitivity was 100% (CI = 89%-100%), specificity 95% (CI = 81%-99%), PPV 94% (CI = 77%-99%), and the NPV was 100% (CI = 92%-100%). The area under the ROC curve for CFD was 95%, compared to 83% for MRA (p = 0.0005). We conclude that the low specificity of MRA precludes its use as the definitive imaging modality for carotid stenosis. The 93% specificity of CFD alone warrants its consideration as a definitive carotid imaging study. By ROC curve analysis, CFD offers superior accuracy to MRA. Our data support noninvasive preoperative carotid imaging for detecting a threshold stenosis of > or =60% whether CFD is used alone, or in combination with the selective use of MRA.