118 chapter 2 Best-case and worst-case scenarios Sound meta-analysis of the performance of BRAF mutation analysis was hindered by a high rate of missing histopathology (36.0%, 2,185/6,071), especially in BRAF mutation-negative indeterminate nodules (37.9%, 2,116/5,577). This could have caused an underestimation of the tests’ specificity (i.e., relatively more BRAF mutation-positive nodules with benign histology (false-positives) were operated on than BRAF mutation-negative nodules with benign histology (true-negatives)) and overestimation of the sensitivity (i.e., by missing malignancies in the unoperated BRAF mutationnegative nodules). To overcome these issues, best-case and worst-case scenarios were constructed, including all 6,071 indeterminate thyroid nodules with a conclusive mutation analysis. In the bestcase scenario, the test result was always correct in the nodules with a conclusive BRAF mutation analysis but no histopathological follow-up. In the hypothetical best-case scenario, the estimated pooled sensitivity, specificity, positive and negative LR would be 20.4% (95% CI: 13.6%-29.3%), 99.9% (95% CI: 99.7%-100%), 201.03 (95% CI: 67.64-597.44) and 0.80 (95% CI: 0.72-0.88), respectively. In the hypothetical worst-case scenario, the results of BRAF mutation analysis were considered false in all unresected nodules. The estimated pooled sensitivity, specificity, positive and negative LR would be 8.8% (95% CI: 5.8%-13.3%), 100% (95% CI: 99.8%-100%), 14,962.20 (95% CI: 44.69-5,009,532) and 0.91 (95% CI: 0.88-0.95), respectively (Figure 4). This shows that, regardless whether its true performance is closer to the best-case or to the worst-case scenario, the specificity of BRAF mutation analysis remains nearperfect but its sensitivity inadequate. Figure 11. SROC curve of BRAF mutation analysis – Bethesda IV Summary receiver operating characteristic plot showing sensitivity versus 1-specificity of BRAF mutation analysis in Bethesda IV thyroid nodules with available histopathology. AUC = 0.96 (95% CI: 0.93-0.97). AUC, area under the curve; HSROC, hierarchical summary receiver operating characteristic.
RkJQdWJsaXNoZXIy MTk4NDMw