Analysis of Doppler waveform morphology for features of the tardus-parvus phenomenon has been promoted as a useful and accurate means for detecting renal artery stenosis. The purpose of this study was to examine and quantify the interobserver agreement of such an analysis and to determine if interobserver differences limit the value of this approach for predicting renal artery stenosis.

Four observers independently categorized renal artery waveforms of 47 patients (94 kidneys) clinically selected for renovascular hypertension. Waveforms were classified into five categories based on the presence and severity of tardus-parvus changes in the systolic upstroke and early systolic peak. This categorization was then compared with angiographic findings, and the results were analyzed with receiver-operating-characteristic curves. Kappa statistics and agreement tables were computed to evaluate interobserver agreement.

Interobserver agreement in the waveform analysis for the four interpreters was statistically significant (p < 0.001). The receiver-operating-characteristic areas produced by the observers indicated, however, that such waveform classification was not strongly predictive of renal artery stenosis.

We conclude that substantial agreement in the interpretation of waveform morphology can be obtained between independent observers, and that such differences that do exist do not preclude the use of the pattern-recognition approach to waveform analysis. Even so, the specific application of this strategy to the waveform contours of early systole was not successful in predicting the presence or severity of renal artery stenosis.