Endocrine Abstracts

Introduction: Papillary thyroid cancer (PTC) is an indolent tumor with low malignancy potential. However, occult synchronous cervical lymph node metastases (CLNM), which may lead to a high risk of local recurrence, are still present in 20–40% of patients. The low diagnostic efficiency (< 47%) of the neck ultrasound in central lymph node metastases detection leads to the development of alternative ways of occult CLNM prediction in patients with PTC. The objective is to develop a prognostic model based on preoperative clinical and ultrasound predictors for the assessment of the papillary thyroid cancer local metastasis risk.

Materials and methods: The perioperative data of 301 patients were analyzed: 117 patients with CLNM, diagnosed after surgery; 184 patients - without CLNM. Five factors were determined as the most significant CLNM predictors: 1) subcapsular location; 2) tumor size; 3) ill-defined margin; 4) microcalcifications; 5) age. The prognostic model was created on the basis of binary logistic regression using the statistical analysis program StatPlus, version 7 (AnalystSoft Inc.). An MS Excel template was created to automate the calculation process.

Results: Regression coefficients calculations were performed:

The probability of metastases detection was calculated by the formula: P = 1/ (1 - e^-y), where e - the Euler`s number, y - obtained by regression equation, i.e.: y = –2.009 – 0.039*X1 + 0.114*X2 + 1.033*X3 + 1.215*X4 + 0.590*X5 X are binary variables (presence or absence of the feature): X <sub>1</sub> - age (years), X <sub>2</sub> - tumor size (threshold value is ≥ 10 mm), X <sub>3</sub> - subcapsular location, X <sub>4</sub> - ill-defined margin, and X <sub>5</sub> - microcalcifications. The resulting prognostic model had sensitivity – 61.5 % (95% CI 52.1-70.4); specificity – 83.7 % (95% CI 77.6-88.7); diagnostic efficiency – 75.1 % (95% CI 69.8-79.9). The model`s quality was assessed (n =50): sensitivity – 53.7 %, specificity – 77.1 %, and diagnostic efficiency - 70.0%. The values of the selected indicators were within the calculated 95% CI. The values of the selected indicators were generally within the calculated 95% CI, however, the model needs further improvement on a larger sample size.

Conclusions: The practical application of the suggested prognostic model may improve the outcomes of PTC management by preoperative predicting the risk of CLNM presence with an accuracy of 75.1%.