*2.5. Statistical Analysis*

Gal-3 IHC score data were analyzed using IBM SPSS Statistics version24 (SPSS, Chicago, IL, USA, http://www.ibm.com/analytics/us/en/technology/spss/ (accessed on 2 March 2020). Descriptive statistic was used to describe the patient population and inferential statistic was used to test the Gal-3 expression differences between the groups (one-way ANOVA). Gal-3 positivity scores served as the predictor variable and invasion served as the primary outcome variable (invasiveness). Diagnostic accuracy was assessed by plotting receiver operator characteristics (ROC) curve. The cut off points for Gal-3 positivity expression and area under the curve (AUC) were calculated. The optimal cutoff value was chosen as the threshold that maximized the AUC. The clinical utility of biomarkers was assessed using AUC, positive predictive value (PPV), negative predictive value (NPV) for invasiveness. The association between cytoplasmic Gal-3 positivity and invasiveness was examined by odds ratio (OR) using cross-tabulation. For the accuracy assessment of positive cytoplasmic Gal-3 expression in distinguishing NIFTP and invasive EFVPTC, we calculated the likelihood ratios (LR): (LR+ = sensitivity/(1-specificity)) and

(LR− = ((1-sensitivity)/specificity). The results were considered statistically significant at *p* ≤ 0.05.
