Accuracy of carotid Doppler peak frequency analysis: Results determined by receiver operating characteristic curves and likelihood ratios

To determine the accuracy of peak Doppler frequency measurements in the diagnosis of extracranial carotid arterial disease and to apply the results to the assessment of the individual patient, 397 studies of continuous-wave Doppler frequency analysis were compared with the results of arteriography. The diagnostic accuracy of the measurement of peak frequency was determined in two ways. First, receiver operator characteristic curves were calculated for different threshold levels of peak frequency. The results showed that the measurement of the peak frequency can be related in only an approximate way to the severity of the arteriographic stenosis because of the uncertainty of the probe-to-vessel angle. Nevertheless, the overall maximum diagnostic accuracy was approximately 91% (Kappa = 0.82 +/- 0.05). Second, likelihood ratios were calculated to take into account the degree of abnormality of the peak frequency measurements. The probability of carotid disease increased as the peak frequency increased, giving likelihood ratio ranges from 0.03 to 25.71. To be able to apply the peak frequency measurements to the assessment of the individual patient in the vascular laboratory, our results were used to construct tables that show the posttest probability of carotid disease based on the clinically estimated pretest probability of disease and the measured peak frequency. It is concluded that the measurement of peak Doppler frequency is an adjunct in the diagnosis of carotid artery disease.