Wiseman Sam M, Melck Adrienne, Masoudi Hamid, Ghaidi Fariba, Goldstein Lynn, Gown Allen, Jones Steven J M, Griffith Obi L
Department of Surgery, St. Paul's Hospital, University of British Columbia, C303-1081 Burrard Street, Vancouver, BC, Canada.
Ann Surg Oncol. 2008 Oct;15(10):2811-26. doi: 10.1245/s10434-008-0034-8. Epub 2008 Jul 9.
Currently, a large proportion of individuals undergo thyroidectomy as a diagnostic procedure for cancer. The objective of this work was to evaluate the molecular phenotype of differentiated thyroid cancer (DTC) and benign thyroid lesions to identify molecular markers that allow for accurate thyroid cancer diagnosis.
Tissue microarrays consisting of 100 benign and 105 malignant thyroid lesions, plus 24 lymph node samples, were stained for a panel of 57 molecular markers. Significant associations between marker staining and tumor pathology (DTC versus benign) were determined using contingency table and Mann-Whitney U (MU) tests. A Random Forests classifier algorithm was also used to identify useful/important molecular classifiers.
Of the 57 diagnostic markers evaluated 35 (61%) were significantly associated with a DTC diagnosis after multiple testing correction. Of these, in DTC compared with benign thyroid tumors, 8 markers were downregulated and 27 upregulated. The most significant markers for DTC diagnosis were: Galectin-3, Cytokeratin 19, Vascular Endothelial Growth Factor, Androgen Receptor, p16, Aurora-A, and HBME-1. Using the entire molecular marker panel, a Random Forests algorithm was able to classify tumors as DTC or benign with an estimated sensitivity of 87.9%, specificity of 94.0%, and an accuracy of 91.0%.
Evaluation of the DTC and benign thyroid tumor molecular phenotype has allowed for identification of a marker panel, composed of both established and novel markers, useful for thyroid cancer diagnosis. These results suggest that further study of the molecular profile of thyroid tumors is warranted, and a diagnostic molecular marker panel may potentially improve patient selection for thyroid surgery.
目前，很大一部分人接受甲状腺切除术作为癌症的诊断手段。本研究的目的是评估分化型甲状腺癌（DTC）和良性甲状腺病变的分子表型，以确定能够实现甲状腺癌准确诊断的分子标志物。
对包含100个良性和105个恶性甲状腺病变以及24个淋巴结样本的组织微阵列进行57种分子标志物的染色。使用列联表和曼-惠特尼U（MU）检验确定标志物染色与肿瘤病理（DTC与良性）之间的显著关联。还使用随机森林分类算法来识别有用/重要的分子分类器。
在评估的57种诊断标志物中，经过多重检验校正后，35种（61%）与DTC诊断显著相关。其中，与良性甲状腺肿瘤相比，DTC中有8种标志物下调，27种上调。DTC诊断最显著的标志物为：半乳糖凝集素-3、细胞角蛋白19、血管内皮生长因子、雄激素受体、p16、极光激酶A和人甲状腺髓样癌单克隆抗体-1。使用整个分子标志物面板，随机森林算法能够将肿瘤分类为DTC或良性，估计灵敏度为87.9%，特异性为94.0%，准确率为91.0%。
对DTC和良性甲状腺肿瘤分子表型的评估已确定了一个由既定和新型标志物组成的标志物面板，可用于甲状腺癌诊断。这些结果表明有必要进一步研究甲状腺肿瘤的分子谱，并且诊断性分子标志物面板可能会潜在地改善甲状腺手术的患者选择。
