Accuracy of Frozen Section for Thyroid Nodules with Atypia of Undetermined Significance

Abstract

Introduction: Thyroid cytology with atypia of undetermined significance (AUS/Bethesda III) carries 15–30% malignancy risk. Frozen section (FS) during surgery may guide extent of resection, but its accuracy in AUS nodules is debated.
Methods: Systematic review of 12 studies (n=1,078 patients with Bethesda III nodules undergoing lobectomy/thyroidectomy). FS diagnosis (benign, malignant, deferred) compared to permanent section (PS) as reference. Outcomes: sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) of FS for malignancy. Subgroup analysis by nodule size (>4 cm vs ≤4 cm).
Results: FS sensitivity was 58% (95% CI 48–67%) - meaning 42% of malignancies missed. Specificity was 98% (95–99%). PPV was 92%, NPV 82%. Missed malignancies were mostly follicular variant papillary carcinoma or microcarcinomas. FS performed better for nodules >4 cm (sensitivity 71% vs 49%). Deferred diagnosis (FS indeterminate) occurred in 18%.
Discussion: Frozen section has low sensitivity for AUS nodules, potentially leading to completion thyroidectomy in two stages. Molecular testing (ThyroSeq, Afirma) may be superior.
Conclusion: Frozen section is not recommended for routine intraoperative guidance in Bethesda III thyroid nodules due to low sensitivity.