Development and validation of a preoperative prediction model for follicular thyroid carcinoma

Abstract Objective The low pre‐ and intraoperative diagnostic rates in follicular thyroid carcinoma (FTC) often lead to inadequate surgical resection and necessitate further completion surgery. Therefore, the preoperative prediction of FTC in thyroid nodules is essential. Design and Patient Patients were categorized into two data sets: the modelling data set, which included 3649 patients admitted to our centre between January 2014 and December 2016, and the validation data set, which included 1253 patients admitted between January and December 2017. Patient data from the FTC and non‐FTC groups were initially included in a modelling data set to establish a preoperative prediction model. This model was subsequently employed in a validation data set for external validation of the predictive value. The positivity rate for FTC predicted by the model was compared with that of the intraoperative frozen sections. Results The preoperative serum thyroglobulin level, nodule diameter, calcification status, solidity and blood supply were selected as predictors for the model. The regression equation was as follows: Y = 0.010 × (thyroglobulin level) + 0.556 × (nodule diameter) + 0.675 × (calcification status) + 2.355 × (nodule component) + 1.072*(blood flow) − 9.787. The model positively predicted FTC at values of Y ≥ −4.11. The accuracy, sensitivity, specificity, positive likelihood ratio and negative likelihood ratio of the prediction model were 89.2%, 90.2%, 87.7%, 39.2 and 0.11, respectively. External validation of the model demonstrated acceptable results. The positive prediction rate of the model was 90.7% (78/86), which was significantly higher than that of the intraoperative frozen sections (10.5% [9/86]; P < 0.0001). Conclusions We successfully established and validated a simple and reliable preoperative prediction model for FTC using the preoperative thyroglobulin level and ultrasonographic features of the thyroid nodules. This model may improve the preoperative evaluation of FTC in clinical settings and facilitate the development of a reasonable surgical programme for FTC.