*3.3. Cytoplasmic Gal-3 Predictive Value for NIFTP versus Invasive EFVPTC*

The accuracy of cytoplasmic Gal-3 expression in distinguishing invasive EFVPTC from NIFTP and benign thyroid nodules was determined by the ROC curve analysis. Cytoplasmic Gal-3 emerged as the strongest predictor of invasive EFVPTC in comparison with nonmalignant tissues (AUC = 0.90, (CI 0.83–0.97), *p* < 0.001) thereby underscoring its potential clinical applicability (Table 5 and Figure S2). With a positive cut off value 3.71 (Table S1), cytoplasmic Gal-3 expression in EFVPTC versus in NIFTP showed a sensitivity of 75.6% (95% CI: 63.9–87.2%), specificity 80.5% (95% CI: 68.6–92.4%), positive predictive

value (PPV) 81.0% (95% CI: 69.6–92.3%) and negative predictive value (NPV) 75.0% (95% CI: 62.8–87.2%) (Table 5).

**Table 5.** The clinical utility of biomarkers was assessed using sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and area under the curve (AUC) between benign or NIFTP vs. invasive EFVPTC group.


The cross tabulation analysis of diagnostic odds ratio (OR) showed the positive cytoplasmic Gal-3 expression in invasive EFVPTC was 13 times higher risk of having adverse outcome compared to indolent NIFTP (OR = 12.75, 95% CI: 5.88–49.14). The positive likelihood ratio (LR+ = 3.87, 95% CI: 2.21–7.97) indicated the probability in diagnosing as malignant tumors was increased 2.1 times more in specimens with the positive cytoplasmic Gal-3 expression. However, the negative likelihood ratio (LR- = 0.30, 95% CI: 0.14–0.45) showed the probability of malignancy was decreased by 30% in specimens with the low cytoplasmic Gal-3 expression.
