196 chapter 2 Immunocytochemistry Our systematic literature search yielded 19 articles that investigate the role of immunocytochemistry in indeterminate thyroid nodules, including 1,317 nodules in total, and investigating a great variety of ICC markers. As 11 studies investigated the combined use of multiple markers, one table is presented with baseline characteristics for all studies (Table 44). The investigated ICC markers included: galectin-3 in 12 studies [82, 202-204, 208, 212-215, 217, 219, 228]; HBME-1 in seven [203, 212, 214, 217-219, 224]; cytokeratin 19 (CK-19) in three [212, 218, 220] and Ki-67 in two [218, 223]. Thyroid peroxidase (TPO), keratan sulphate, laminin, cyclin D1, cyclin D3, emerin, P27, TROP-2, and RET proto-oncogene were each investigated by a single study [203, 212, 214, 219, 223, 224, 226, 227]. ICC was applied to formalin-fixed paraffin-embedded (FFPE) cell blocks in nine studies [82, 202, 208, 212, 214, 220, 226-228]; liquid based cytology in five [203, 217, 218, 224, 355]; cytology smears in five [203, 204, 213, 219, 223] and was unspecified in one [215]. The individual ICC markers are discussed in the following sections. Galectin-3 Meta-analysis was performed of the twelve studies that performed ICC using galectin-3, including a total of 1,052 indeterminate thyroid nodules [82, 202-204, 208, 212-215, 217, 219, 228]. The average malignancy rate was 36.4% (383/1,052). The 33 FT-UMP lesions disregarded by Bartolazzi et al. were reclassified as malignant in accordance with our definitions [202, 215]. Sensitivity and specificity of galectin-3 to detect thyroid carcinoma in indeterminate thyroid cytology ranged from 0% to 92% and from 68% to 100%, respectively. A positive galectin-3 stain was found in 33.4% (351/1,052) nodules. Histopathology was available in 97.3% (1,016/1,044) of the cases with a conclusive ICC result. Presence of study heterogeneity is suggested by an I2 of 67.8% for sensitivity and 72.4% for specificity. The estimated pooled sensitivity, specificity, positive LR and negative LR are 73.9% (95% CI: 62.6%-82.8%), 90.7% (95% CI: 82.7%-95.2%), 7.97 (95% CI: 4.25-14.95) and 0.29 (95% CI: 0.20-0.42), respectively. The AUC is 0.90 (95% CI: 0.87-0.92) (Table 45, Figure 53 and Figure 54). For a given prevalence of malignancy of 15%, 25% or 40%, these results correspond to an estimated PPV and NPV of 58.4% (95% CI: 42.9%-72.5%) and 95.2% (95% CI: 93.1%-96.6%), 72.7% (95% CI: 58.6%-83.3%) and 91.3% (95% CI: 87.7%-93.8%), or 84.2% (95% CI: 73.9%-90.9%) and 83.9% (95% CI: 78.1%-88.4%), respectively (Figure 55). Subgroup analyses for Bethesda III and IV nodules could not be performed. Best-case and worst-case scenarios are nearly identical as histopathology from just 3.6% of nodules with a conclusive index test was missing (Figure 4).
RkJQdWJsaXNoZXIy MTk4NDMw