237 Diagnostic utility of molecular and imaging biomarkers 2 Elastosonography Our systematic literature search yielded 14 articles on the diagnostic value of USE for thyroid nodules with an indeterminate FNAC result (Table 54). Most studies excluded nodules that were hard to diagnose on USE due to technical limitations: nodules with egg shell calcifications or mainly cystic content, and multiple coalescent nodules [248, 253, 255, 257, 259, 260, 287, 352, 353, 359]. In total, 1,389 indeterminate thyroid nodules were included. The average malignancy rate was 28.5% (396/1,389), including 188 PTC (51%), 99 FVPTC (25%), 62 FTC (16%), eight FTC-OV (2%), two MTC (0.5%), three other (0.8%) and 34 unspecified types of thyroid carcinoma (8.6%) (Table 55). Ten studies performed qualitative USE, nine of which used one of the known color-coded scoring systems [248, 251, 253, 259, 260, 287, 352, 353, 359]. Piccardo et al. used a dichotomous color-coded system, distinguishing elastic from non-elastic nodules [58]. Five studies performed semi-quantitative measurements of the strain index [253, 255, 257, 278, 287], and one recent study performed quantitative shear-wave USE [57]. Two of these studies performed both a qualitative and a semi-quantitative analysis [253, 287]. Separate meta-analyses for qualitative and semi-quantitative USE were performed. Subgroup analyses for Bethesda III and IV nodules were not performed, as most studies included indeterminate or Thy3 cytology without further specification. Best-case and worst-case scenarios are identical to the presented results, as histology was available in all cases. Meta-analysis was performed of the 10 studies on qualitative USE, involving 865 indeterminate thyroid nodules. The studies applied different qualitative elasticity score systems (Table 54). First, pooled diagnostic indexes were assessed using the elasticity thresholds in the same way that they were applied in the original articles. There were 356 (41.1%) nodules with a positive test (low elasticity) and 509 (58.8%) with a negative test (high elasticity). The average malignancy rate was 28.7% (248/865). Estimated pooled sensitivity, specificity, positive LR and negative LR are 80.0% (95% CI: 68.5%-88.0%), 81.2% (95% CI: 52.9%-94.4%), 4.27 (95% CI: 1.37-13.3) and 0.25 (95% CI: 0.14-0.45), respectively. The AUC is 0.85 (95% CI: 0.82-0.88) (Table 56 and Figure 75 and Figure 76). For a given prevalence of malignancy of 15%, 25% or 40%, these results correspond to an estimated PPV and NPV of 42.9% (95% CI: 19.5%- 70.0%) and 95.8% (95% CI: 92.7%-97.7%), 58.7% (95% CI: 31.4%-81.5%) and 92.4% (95% CI: 87.0%- 95.7%), or 74.0% (95% CI: 47.8%-89.8%) and 85.9% (95% CI: 77.0%-91.7%), respectively (Figure 77). As suggested by Nell et al., we also analysed the diagnostic performance of qualitative USE using a different elasticity cut-off value. Regardless of the initial score system, we defined an elasticity score of 1 (entirely soft nodule) as benign (negative index test), and all higher scores as suspicious (positive index test) [286]. The individual patient data necessary for this analysis could be retrieved for 713 nodules from six studies [248, 251, 253, 260, 287, 353]. The study by Piccardo et al. was excluded from this analysis because of their dichotomous scoring system, which did not differentiate between entirely and partly soft nodules. In total, 142 (19.9%) nodules had a qualitative USE score of 1. Estimated pooled sensitivity, specificity, positive LR and negative LR are 98.5%
RkJQdWJsaXNoZXIy MTk4NDMw