Methods: From January 1997 to January 2000, a prospective multicenter study was conducted that was approved by the medical ethical committee. After informed consent was obtained patients with symptoms of amaurosis fugax, TIA, or minor stroke were included, who underwent duplex ultrasonography and digital subtraction angiography. Selective ipsilateral carotid angiograms were obtained in at least three planes. The PSV was measured in the proximal part of the internal carotid artery. Arteries with near or total occlusions on duplex ultrasonography were excluded because the PSV is not reliably measurable in these arteries. Receiver operating characteristic (ROC) curves were constructed for the diagnosis of either a 70-99% or a 50-99% stenosis. Based on the lifetime costs and quality-adjusted life-years, obtained from a cost-effectiveness analysis, and the prevalence of disease, the optimal likelihood ratio was calculated. Finally, the associated optimal sensitivity, specificity and threshold PSV were derived from the ROC-curves.

Results: In the clinical study 350 patients were included. In total 236 non-occluded arteries were available for the ROC analysis. For the diagnosis of a 70-99% stenosis the optimal likelihood ratio was 0.21, which was associated with a threshold PSV of 220 cm/s, a sensitivity of 97% (127/131; 95% CI 94-100%), and specificity of 48% (50/105; 95% CI 38-57%). For the diagnosis of a 50-99% stenosis the optimal likelihood ratio was 0.38, which was associated with a threshold PSV of 180 cm/s, a sensitivity of 95% (182/191; 95% CI 92-98%), and specificity of 69% (31/45; 95% CI 55-82%).

Conclusions: Based on the lifetime consequences of diagnostic testing and subsequent treatment, the optimal threshold PSV for the diagnosis of a 70-99% carotid artery stenosis in patients with amaurosis fugax, TIA, or minor stroke is 220 cm/s and for the diagnosis of a 50-99% stenosis the optimal threshold PSV is 180 cm/s.