The diagnostic value of colour duplex ultrasound for symptomatic carotid stenosis in clinical practice

We assessed the accuracy of colour duplex ultrasound for the detection of severe (70–99 %) symptomatic carotid stenosis in a clinical setting, in order to assess whether it could make carotid angiography unnecessary. In 152 patients with a transient ischaemic attack or non-disabling ischaemic stroke in the carotid distribution, we compared the degree of colour duplex ultrasound stenosis with angiographic stenosis by receiver-operating-characteristic analysis. The angiograms were evaluated by blinded observers, and compared with routine reports of the colour duplex examination. We computed the sensitivity and specificity of colour duplex, and the number of angiograms and sonographic studies needed to prevent one stroke within 3 years, taking into account the risks of angiography, and the risks and efficacy of endarterectomy. The estimates were adjusted for nonverification bias. We found 34 patients (22 %) with a severe (70–99 %) symptomatic carotid stenosis. In 16 patients (11 %) the symptomatic artery was occluded. The sensitivity and specificity of duplex ultrasound were 76 % and 85 %, respectively. The number of patients needed to undergo angiography to prevent one stroke was reduced from almost 200 to 33, when colour duplex was used as a preoperative examination. After adjustment for the effects of nonverification, the sensitivity dropped to 58 % and the number of duplex studies needed to prevent one stroke would double. The number of angiograms needed after positive duplex sonography would be virtually unaffected. Were colour duplex sonography to have been the sole preoperative investigation, the number needed to diagnose to prevent one stroke within 3 years would be approximately 350, more than twice as many as with the combined diagnostic strategy. The diagnostic accuracy of colour duplex sonography in clinical practice seems less impressive than previous studies have suggested, but it remains an effective way to select patients for angiography. Its use as a single preoperative assessment cannot be recommended. Received: 24 March 1998 Accepted: 21 May 1998