Endocrine Abstracts

Background: In 2014 a new Italian classification system of thyroid cytology divided the indeterminate TIR3 category in two groups, TIR3A and TIR3B, aiming to reduce unnecessary thyroidectomies. The reported risk of malignancy is: TIR3A: <10%, suggesting follow-up and possible FNA repetition after 6 months, and TIR3B: 15-30%, recommending surgery.

Objectives: We aimed to: evaluate the histological prevalence of malignancy in TIR3A and TIR3B nodules in our center; investigate whether oxyphil cells in TIR3B samples correlated with benignity; assess whether cytological ThinPrep versus conventional smear preparation affects the cytological report; estimate the association between clinical and ultrasound (US) features with malignancy.

Methods: We performed a retrospective analysis of patients who received fine needle aspiration (FNA) from 05/2014 to 07/2017 at Fondazione Policlinico Gemelli, Rome. We included 139 TIR3A and 162 TIR3B nodules who underwent surgery, to compare cytology and histology. Samples were obtained with capillarity technique and prepared on conventional smears or ThinPrep slides. Clinical and US data reported to be associated with malignancy were collected. Differences in rates of malignancy were evaluated with Fisher’s test. Logistic regression was performed to identify predictors of malignancy.

Results: Malignancy was reported in 12.2% (n=17) of TIR3A and 27.1% (n=44) of TIR3B nodules, with no significant difference with literature data. In TIR3B subgroup, 83 cytological samples showed oxyphil cells, with a malignancy rate of 10.8% (n=9), significantly lower than overall TIR3B (P<0.01). 66 TIR3A and 109 TIR3B FNA were prepared with ThinPrep technique, and 73 TIR3A and 53 TIR3B with conventional smears, with no difference in malignancy risks. Logistic analysis showed irregular margins to be predictive of malignancy in both TIR3A (OR 10.75, 95%CI 2.25-51.37) and TIR3B (OR 6.80, 95%CI 1.94-24.01). Other predictors were family history and microcalcifications for TIR3A, and vascularity, hypoechogenicity and, secondarily, microcalcifications for TIR3B (P<0.05). The predictive power of the logistic model increased when considering those features concurrently.

Conclusion: In a representative cohort of indeterminate nodules, malignancy rate of TIR3A and TIR3B nodules was overall comparable to the one reported in the Italian classification system. However, TIR3B with oxyphil cells had a malignancy rate comparable to TIR3A, advocating for careful search of this feature. US findings of irregular margins and, to a lesser extent, microcalcifications and hypoechogenicity, proved to be associated with malignancy. Some vascularity patterns were also associated with malignancy but they remained unreliable features because of their dubious interpretation. FNA preparation technique had no effect on malignancy risk.