Doppler velocity criteria based on receiver operating characteristic analysis for the detection of threshold carotid...

Duplex scanning is an accepted method for noninvasive evaluation of carotid stenosis. However, the ultrasound criteria used for the detection of threshold stenoses vary widely between laboratories, and quality assurance measures to allow adjustment of criteria are often lacking. This study was completed using receiver operating characteristic (ROC) analysis to determine Doppler velocity criteria for threshold carotid stenoses, compared to an accepted standard, and to demonstrate methods to allow adjustment of criteria.

The study cohort included 134 patients who had carotid endarterectomy. Ultrasound and arteriographic data were collected for both the operated and nonoperated sides. Each carotid artery was treated as an independent case in the final analysis. Angiograms were used as the gold standard in ROC analysis to determine the Doppler velocity criteria for the detection of different threshold stenoses.

The ROC analysis results showed that for the detection of 70% stenosis, the best Doppler systolic criterion was 200 cm/s (sensitivity 93.6%, specificity 71.7%, area under the curve [AUC] 87.6%), the best diastolic criterion was 65 cm/s (sensitivity 85.1%, specificity 74.6%, AUC 84.3%), and the best criterion of carotid ratio (CR) (internal carotid artery systolic velocity/common carotid artery systolic velocity) was 3.0 (sensitivity 78.7%, specificity 75.4%, AUC 81.3%). For 50% stenosis, the best systolic criterion was 140 cm/s (sensitivity 90.3%, specificity 95.2%, AUC 97.0%), the best diastolic criterion was 60 cm/s (sensitivity 98.6%, specificity 77.8%, AUC 92.1%), and the best criterion of CR was 2.5 (sensitivity 93.1%, specificity 72.0%, AUC 89.0%).

This study showed that duplex scanning is able to detect threshold carotid stenoses. For the best performance, each laboratory should have its own criteria; however, the criteria provided here could be a helpful reference to those laboratories that have not yet established their own criteria. Most important, this study provides an example of how to evaluate the performance criteria, how to modify them, how such changes can affect performance, and how performance can be modified depending on the goals of the laboratory.