Reproducibility in ultrasonic characterization of carotid plaques

Background and Purpose: Echolucent carotid plaques compared with echogenic plaques could carry a significant risk of transient ischemic attacks and strokes, but the reproducibility of new ultrasonic methods has not yet been proved. The objective was to evaluate interobserver and intraobserver agreement in characterizing the carotid plaques studied by both B mode imaging and color Doppler imaging, which is the only ultrasonic method available for recognizing anechoic lesions. Methods: Fifty-three carotid plaques greater than 40% in diameter were selected from four centers and simultaneously analyzed by 9 observers. Five types of plaques were defined by their echo structure: class I = uniformly anechogenic, class II = predominantly hypoechogenic with >50% hypoechogenic area, class III = predominantly echogenic with >50% echogenic area, class IV = uniformly echogenic, class V = unclassified plaques. The luminal surface was characterized as either 1 = regular, 2 = recess of more than 2 mm in depth and width, or 3 = unclassified. Agreement of these variables was calculated by using the kappa index, agreement proportion and an intraclass correlation coefficient. Results: Interobserver reproducibility was only fair for type I (kappa = 0.47) and for the luminal surface (class 1, kappa = 0.52 and class 2, kappa = 0.41). Agreement proportion was 0.51 in hypoechoic plaques and 0.64 in the determination of the regular surface. Mean intraobserver agreement was fair (kappa = 0.47 +/- 0.1) for plaque echogenicity to good (kappa = 0.63 +/- 0.19) for surface. Conclusion: This study shows that the semiquantitative classification, first developed by Gray-Weale, then by Nicolaides, could be improved, thus giving rise to a new outlook in the debated field of ulcerations.