Ultrasound and demographic predictors of papillary thyroid cancer local metastasis at the preoperative stage: a predictive model

Abstract

Background. Due to the low sensitivity and specificity of neck ultrasound in the evaluation of the central lymph nodes, it is necessary to find alternative ways to predict central lymph node metastases in patients with papillary thyroid cancer (PTC). The purpose of the study is to develop a prognostic model for assessing the risk of local metastasis of papillary thyroid cancer based on preoperative ultrasound and demographic predictors. Materials and methods. A retrospective monocenter cohort study was conducted. The data of 401 patients who underwent surgery for PTC were processed. The main group included 179 patients in whom central lymph node metastases were detected during histopathological examination. The comparison group consisted of 222 patients without metastases according to the results of histopathological examination. When conducting the analysis, the following signs were considered as risk factors: subcapsular location of the tumor; size; blurred boundaries; the presence of calcification areas (microcalcifications); age of a patient. Inclusion criteria: PTC that was cytologically confirmed at the preoperative stage; surgeries (thyroidectomy/hemithyroidectomy and central neck lymph node dissection). Exclusion criteria: history of neck surgery; multifocal lesions of the thyroid gland. Results. The following logistic equation was obtained using the StatPlus program: y = –1.839 – 0.037 × X1 + 0.097 × X2 + 1.123 × X3 + 1.198 × X4 + + 0.692 × X5, where X1, X2, X3, X4, X5 are age (years), size (mm), subcapsular location (presence), blurred boundaries (presence), calcifications (presence), respectively. The obtained prognostic model provided the following operational characteristics when specifying the threshold value for P, which is equal to 0.44 (the decisive rule X > T): sensitivity — 75.4 %, specificity — 79.3 %, diagnostic efficiency — 75.1 %. To assess the quality of the model, the receiver operator characteristic (ROC) analysis was performed. The following data were obtained: area under the ROC curve = 0.797 (95% confidence interval: 0.753–0.841), which corresponds to the good quality of the model, and Youden index J = 0.5155. Checking the quality of the model on a control group of 100 people gave the following results: sensitivity — 72.2 %, specificity — 76.1 %, diagnostic efficiency — 75.8 %. Conclusions. The application of a prognostic model based on ultrasound data and the patient’s age makes it possible to predict the presence of local metastases of the PTC at the preoperative stage with a diagnostic efficiency of 75.8 %. No prognostic model gives 100% prediction accuracy. When choosing the treatment, additional influencing factors (history of radiation exposure, burdened family history, etc.) for a specific patient should be considered.