Core needle biopsy (CNB) is now more frequently used for the preoperative diagnosis of thyroid nodules. Based on morphology alone, 5–20% of CNB samples cannot be determined as malignant or benign. Compared to fine-needle biopsy (FNB), samples collected by CNB are more accessible for various tests. Therefore, studying biomarkers’ application in distinguishing uncertain CNB samples of thyroid nodules is a practical need. Patients of thyroid nodules with both CNB and matched resected specimens were reviewed. Cases classified as indeterminate lesions, follicular neoplasms, and suspicious for malignancy were retrieved. All CNB samples were stained by immunohistochemistry (IHC) using antibodies against CK19, galectin-3, HBME-1, and CD56 and detected by next-generation sequencing (NGS) using an OncoAim® thyroid cancer multigene assay kit (Singlera Genomics) that detected 26 genes. Taking the resected specimens’ classification as the gold standard, the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy of a single biomarker, and various combinations for discriminating malignancy from benignity were calculated. The sensitivity, specificity, PPV, NPV, and accuracy for preoperative malignancy evaluation were as follows. In the cohort of non-follicular-neoplasm-lesions (non-FN-lesion), they were 95.16%, 53.85%, 90.77%, 70.00%, and 88.00% for CK19; 95.16%, 38.46%, 88.06%, 62.50%, and 85.33% for galectin-3; 77.42%, 76.92%, 94.12%, 41.67%, and 58.00% for HBME-1; 66.13%, 100.00%, 100.00%, 38.24%, and 72.00% for CD56; 90.32%, 92.31%, 98.25%, 66.67%, and 90.67% for NGS; and 88.71%, 92.30%, 98.21%, 63.16%, and 89.33% for integrated IHC. In the cohort of follicular neoplasms (FN), they were 30.43%, 77.77%, 77.77%, 30.43%, and 43.75% for CK19; 73.91%, 66.67%, 85.00%, 50.00%, and 71.88% for galectin-3; 26.09%, 88.89%, 85.71%, 32.00%, and 43.75% for HBME-1; 26.09%, 100.00%, 100.00%, 34.62%, and 46.88% for CD56; 52.17%, 88.89%, 92.31%, 42.11%, and 62.50% for NGS; 82.61%, 66.67%, 86.36%, 60.00%, and 78.13% for integrated IHC; and 100%, 66.67%, 88.46%, 100%, and 90.63% for integrated IHC-NGS. The application of biomarkers in distinguishing uncertain CNB samples of thyroid nodules is available and capable. CD56 negative or NGS positive suggests malignancy strongly for both FN and non-FN-lesion, which may be used as a “rule in” tool. The negative predictive value of the integrated IHC and the integrated IHC-NGS implies a high possibility to be benign for non-FN-lesion and FN separately, which can work as a “rule out” tool. Considering the balance of specificity and sensitivity, NGS is the best for non-FN-lesion and the integrated IHC-NGS is the best for FN.

芯针活检（CNB）现在更常用于甲状腺结节的术前诊断。仅根据形态学，5-20% 的 CNB 样本无法确定为恶性或良性。与细针活检 (FNB) 相比，CNB 采集的样本更易于进行各种测试。因此，研究生物标志物在鉴别甲状腺结节不确定CNB样本中的应用是一种现实需要。回顾了具有 CNB 和匹配切除标本的甲状腺结节患者。检索分类为不确定病变、滤泡性肿瘤和可疑恶性肿瘤的病例。所有 CNB 样品均使用针对 CK19、galectin-3、HBME-1、和 CD56，并使用检测到 26 个基因的 OncoAim® 甲状腺癌多基因检测试剂盒（Singlera Genomics）通过下一代测序（NGS）检测。以切除标本的分类为金标准，计算了单个生物标志物鉴别良恶性的敏感性、特异性、阳性预测值（PPV）、阴性预测值（NPV）、准确性以及各种组合。术前恶性肿瘤评估的敏感性、特异性、PPV、NPV和准确性如下。在非滤泡性肿瘤病变（非 FN 病变）队列中，CK19 分别为 95.16%、53.85%、90.77%、70.00% 和 88.00%；半乳糖凝集素 3 为 95.16%、38.46%、88.06%、62.50% 和 85.33%；HBME-1 为 77.42%、76.92%、94.12%、41.67% 和 58.00%；CD56 为 66.13%、100.00%、100.00%、38.24% 和 72.00%；90.32%，92。NGS 为 31%、98.25%、66.67% 和 90.67%；综合 IHC 分别为 88.71%、92.30%、98.21%、63.16% 和 89.33%。在滤泡性肿瘤 (FN) 队列中，CK19 分别为 30.43%、77.77%、77.77%、30.43% 和 43.75%；半乳糖凝集素 3 为 73.91%、66.67%、85.00%、50.00% 和 71.88%；HBME-1 为 26.09%、88.89%、85.71%、32.00% 和 43.75%；CD56 为 26.09%、100.00%、100.00%、34.62% 和 46.88%；NGS 分别为 52.17%、88.89%、92.31%、42.11% 和 62.50%；综合 IHC 分别为 82.61%、66.67%、86.36%、60.00% 和 78.13%；集成 IHC-NGS 为 100%、66.67%、88.46%、100% 和 90.63%。生物标志物在区分不确定的甲状腺结节 CNB 样本中的应用是可用的并且有能力的。CD56 阴性或 NGS 阳性强烈提示 FN 和非 FN 病变均为恶性，可用作“规则”工具。综合 IHC 和综合 IHC-NGS 的阴性预测值意味着非 FN 病变和 FN 分别为良性的可能性很高，可以作为“排除”工具。考虑到特异性和敏感性的平衡，NGS 最适合非 FN 病变，综合 IHC-NGS 最适合 FN。