Non‐Oncocytic Thyroid Follicular Neoplasm: Cytological Atypia and Development of a Score for Malignancy Risk Stratification

Vollständiger Abstract

Worum geht es in dieser Arbeit?

ABSTRACT Background Cytological diagnosis of follicular neoplasm (FN; Bethesda IV) is associated with intermediate malignancy risk. Further stratification based on cytological features may be particularly relevant for non‐oncocytic FN but requires reliable, reproducible assessment approaches. This study evaluated cytological features of non‐oncocytic FN in terms of interobserver agreement and association with malignancy risk, and developed a cytological atypia scoring system for risk stratification. Methods This cross‐sectional study included adult patients who underwent surgery for non‐oncocytic FN. Cytological slides were independently reviewed by five cytopathologists. Fifteen cytological features were assessed; interobserver agreement and feature frequencies across histological outcome groups were evaluated. Results The final cohort included 200 cases: 80 malignant, 80 benign, and 40 low‐risk lesions. Eight cytological features showed acceptable interobserver agreement and different degrees of association with malignancy and were incorporated into the scoring system. The cytological atypia score was higher in malignant than in low‐risk and benign groups ( p < 0.001). ROC analysis yielded an AUC of 0.70 (95% CI, 0.63–0.76). Scores ≥ 4 points were associated with higher malignancy risk than scores of 0–3 points (OR = 3.87; 95% CI, 2.01–7.44; p < 0.001). An additional ≥ 8‐point threshold identified a high‐score subgroup with markedly increased malignancy risk compared with the 0–3‐point subgroup (OR = 9.22; 95% CI, 3.43–24.77; p < 0.001). Cytological atypia did not distinguish low‐risk tumours from benign lesions. Conclusion The most reproducible cytological features associated with malignancy risk were identified, and the proposed scoring system enabled stratification of non‐oncocytic FN by malignancy risk, potentially supporting individualised risk assessment after further validation.