Observer variability in the sonographic evaluation of thyroid nodules

Abstract

Inter- and intraobserver variabilities in the description and diagnostic categorization of sonographic (US) features of thyroid nodules were evaluated. The current study was conducted on 72 malignant nodules and 61 benign nodules. The US findings for each thyroid nodule were analyzed twice at a 6-week interval by five radiologists. The analyses were in accordance with the guidelines proposed by the Thyroid Study Group of the Korean Society of Neuroradiology and Head and Neck Radiology (TSGKSNRHNR). Inter- and intraobserver variabilities were calculated using Cohen's kappa statistics. The sensitivity, specificity, positive-predictive value, and negative-predictive value in the assessment of the diagnostic accuracy using these guidelines were calculated. The interobserver agreement was fair to substantial for US features and categorization. Of the US features of the thyroid nodules, internal content (solid versus cystic) showed substantial agreement (k = 0.64). There was moderate agreement with regard to shape, echogenicity, calcification, and diagnostic categories (k = 0.42, 0.57, 0.55, and 0.55, respectively). There was fair agreement for margin, echotexture, and capsule invasion (k = 0.34, 0.26, and 0.32, respectively). With regard to intraobserver agreement, there was moderate to substantial agreement for all US features except for echotexture and capsule invasion, which showed fair agreement. In particular, there was moderate to almost perfect agreement for the diagnostic category. The sensitivity, specificity, positive-predictive value, and negative-predictive value were 65.3%-81.9%, 60.7%-68.9%, 69.7%-73.8%, and 66.6%-75.5%, respectively. There were high degrees of inter- and intraobserver agreement using the "Guidelines for diagnostic thyroid ultrasonography," of the TSGKSNRHNR in the description and categorization of thyroid nodules.