Risk Prediction Model Of Malignant Partially Cystic Thyroid Nodules

Objective:To develop a multivariate logistic regression model and predict the risk of malignant partially cystic thyroid nodules.Method:Ultrasonographic morphology and vascularity of thyroid nodules. Thyroid Stimulating Hormone(TSH) and clinical information of370patients (476nodules) were analyzed retrospectively. All the selected patients were confirmed with pathological diagnosis. The model was developed to calculate the individual risk and evaluated the predictive index.Results:The largest area under the receiver-operating characteristics curve (AUC) was0.90. When apply the model to the verification group,the accuracy、 sensitivity、specificity、misdiagnosis rate、missed diagnosis rate、negative likelihood ratio(LR-) and positive likelihood ratio(LR+) were96.56%,100%、96.29%、3.41%、0%、0and29.32respectively.Conclusions:The model has been developed based on the ultrasonographic variables, blood analysis and clinical information in order to predict the risk of malignant partially cystic thyroid nodules and the individual risk were calculated. The model had a high accuracy to predict the risk of malignancy. But the model would need to be tested prospectively with a large specimen.