43 Diagnostic utility of molecular and imaging biomarkers 2 before morphological tissue changes are seen and therefore considered to be a part of premalignant changes in carcinogenesis [183]. MicroRNA expression profiles are tissue-specific and can not only identify the tissue of origin, but also the histopathological subtype of the cancer and whether it concerns the primary tumour or a metastasis [186, 189]. MicroRNA expression profiles are similar among the various types of thyroid carcinoma, even though expression levels are often distinctively different [186]. In histopathological studies, PTC was associated with an up to 11- to 19-fold upregulation of miR-146b, miR-221, miR-222, miR-181b, miR-187, and a downregulation of miR-1 and miR-138 compared to healthy thyroid tissue and benign nodules. Upregulation of miR-221, miR-222 and miR-187 was also found in FTC, FTC-OV, poorly differentiated and anaplastic carcinoma [183, 184, 186, 190, 191]. Overexpression of miR-146b-3p, miR-146b-5p and miR-375 was seen in both PTC and FVPTC [190, 192]. Furthermore, expression levels of miR-221 and miR-222 were reported about twice as high in FVPTC as compared to PTC or FTC [190]. Only a few microRNAs were differently expressed between follicular neoplasm and FTC [193]. Follicular adenoma was associated with the expression of miR-200a, whereas high expression of miR-31 was found in Hürthle cell adenoma [186]. FTC is related to the differential expression of miR-146b, miR7-5p, miR-346, miR-197 and miR-21, but results among studies are more heterogeneous [186, 193, 194]. FTC-OV showed an expression pattern slightly similar to FTC, but also distinct overexpression of other microRNAs, such as miR-339, miR-183, miR-197 and miR-885-5p [186, 191]. Accordingly, a diagnostic panel of a carefully selected combination of microRNAs and appropriate expression levels could aid in the preoperative distinction of indeterminate thyroid cytology [195]. Recent meta-analyses struggled to reconcile the studies on microRNA in FNAC, as the investigated set of microRNAs was never identical and individual microRNA performance was infrequently described. In unselected cytology, estimated sensitivity of microRNA expression analysis ranged from 75% to 78% regardless of the investigated set; estimated specificity from 73% to 81% [194-196]. In indeterminate thyroid cytology, different sets of microRNAs were evaluated; only several individual microRNAs were analysed in more than one study. The selected microRNAs were first assessed in a test set of cytological and/or histopathological specimens and a cut-off for their expression level was determined. Subsequently, the significantly up- or downregulated microRNAs were validated in an independent set of (indeterminate) thyroid FNAC samples. Some studies developed a decision model for the validation step [187, 192, 197]. The most promising results were presented by Keutgen et al. [197]. Of the six microRNAs investigated in their test set, miR-21, miR-146b, miR-181a and miR-222 were differentially expressed in malignant nodules with prior indeterminate cytology. The subsequently developed support vector machine model incorporated miR-21, miR-222 and the insignificantly expressed miR-197 and miR-328. Prospective validation in an independent set of 72 indeterminate FNAC samples resulted in 100% sensitivity and 86% specificity. Five of the seven false positives had Hürthle cell cytology; excluding these, raised specificity to 95% [197]. Notably, even though overexpression of miR-146b is often related to thyroid carcinoma, it proved not useful to Keutgen et al. to include in their prediction model [197]. In contrast, Agretti et al. and Shen et al. included miR-146b as the key differentiators in
RkJQdWJsaXNoZXIy MTk4NDMw