Objective: Multiple studies have defined criteria for the selection of thyroid nodules for biopsy. No set of criteria is sufficiently sensitive and specific. The aim of this study is to develop a method for assessing consistency of practice in an ultrasound group and to determine whether a 5-point malignancy rating scale can be used to select patients for biopsy.
Materials and methods: One hundred one nodules (50 benign and 51 malignant) were selected from a thyroid biopsy database. Seven radiologists were educated on evidence-based criteria used to select nodules for biopsy. Using this information, readers graded the likelihood of malignancy using a 5-point malignancy rating scale, where 1 equals the lowest probability of malignancy and 5 equals the highest probability of malignancy, on the basis of overall impression of sonographic findings. Interobserver agreement on biopsy recommendation, reader sensitivity, specificity, and accuracy were determined.
Results: The sensitivity and specificity of biopsy recommendation were 96.1% and 52%, respectively. The misclassification rate was 25.7%, and accuracy was 74.3%. Interobserver agreement on biopsy recommendation was fair to substantial (κ, 0.38-0.69). The proportion of agreement was excellent for malignant nodules (0.88-1.0). The risk of malignancy increased with increasing malignancy rating: 4.3% of nodules with a malignancy rating of 1 were malignant versus 93.4% of those assigned a rating of 5.
Conclusion: Our study illustrates a method to evaluate the standard of practice for thyroid nodule assessment among radiologists within an ultrasound group. Application of a 5-point malignancy rating scale to select nodules for biopsy is feasible and shows good diagnostic accuracy.