Endocrine Abstracts

Introduction: Indeterminate cytology still puzzles clinicians, due to its wide range of oncological risks. The new Italian cytological classification introduced in 2014 by SIAPEC-IAP split this category into TIR3A and TIR3B, according to a different cancer risk of <10% and <30%, respectively. Consequently, TIR3B nodules still have a surgical indication, even if more than half of cases will not result in thyroid cancer. The present study aims to analyse a consecutive series of TIR3B nodules referred to surgery, to identify potential features able to improve the surgical indication.

Methods: A consecutive series of thyroid nodules referred to the Endocrine Unit of Careggi Hospital from 1^st May 2014 to 31^st December 2021 has been considered to build a prediction algorithm. The main clinical, US and cytological features have been collected, with particular regard to the secondary cytological features such as cell configurations, presence of colloid, aggregates, anisonucleosis, nuclear atypia and Hurtle cells. According to Akaike Information Criterion (AIC), a stepwise logistic regression has been performed to find the best prediction algorithm, using the histological outcome as the dependent variable. Thereafter, this algorithm has been verified in a smaller confirmatory sample of consecutive TIR3B diagnosed and referred to surgery from 1^st January 2022 to 31^st June 2022.

Results: Of 502 TIR3B nodules referred to surgery, 451 cases were included in the exploratory analysis. The observed rate of positive histology was higher than expected, with 161 cases (36%) of thyroid cancer (P=0.010). Considering the final outcome, a stepwise logistic regression showed that the presence of colloid (OR=0.18, CI95%:0.04-0.88, P=0.034), age>55 years (OR=0.49, CI95%:0.31-0.77, P=0.002) and a nodule size >18 mm (OR=0.354, CI95%:0.22-0.56, P<0.0001) were favourable prognostic factors, at odds with the presence of chronic autoimmune thyroiditis (OR=1.7, CI95%:1.05-2.9, P=0.032), Hurtle Cells (OR=4.2, CI95%:1.35-14.1, P=0.020), hypoechoic nodules (OR=2.79, CI95%:1.62-4.81, P=0.0002), aggregate cells disposition (OR=4.55, CI95%:1.57-13.2, P=0.005), and anisonucleosis (OR=5.15, CI95%:1.6-166.1, P=0.0049). A unified prediction algorithm was built according to the multivariate results. A score >14.5 showed a sensitivity of 60.1% and a specificity of 76.8% in predicting the unfavourable outcome (AUC=0.748, 95CI:0.699-0.797, P<0.0001), which corresponded to an OR=4.98 (95CI:3.24-7.65, P<0.0001). When applying the same algorithm on a confirmatory sample of 58 TIR3B cytology, a threshold of 14.5 points showed a positive and a negative predictive value of 53% and 80%, respectively.

Conclusions: A new predictive algorithm which considers the main clinical, US and cytological features can significantly improve the oncological stratification of TIR3B cytology.