85 Diagnostic utility of molecular and imaging biomarkers 2 data convergence and caused execution failure of the midas command for subgroup analysis of two different index tests: combination of BRAF and RAS mutation analysis in Bethesda IV nodules, and semi-quantitative USE. Pooled results for these index tests are presented without the parameters that could not be estimated, including heterogeneity parameters, AUC and Bayesian probability plots. Statistical analyses were performed using Stata/MP (version 14.2, StataCorp LP, College Station, TX) and Microsoft Excel (version 14.0.7, Microsoft, 2010)[349]. Best-case and worst-case scenarios The present systematic review and meta-analysis explicitly focuses on the interpretation of the index test results in the light of the availability of an appropriate reference test. Dependent on the index test, histopathological confirmation was unavailable in a substantial part of the patients in many studies. To investigate the potential impact of unavailable histopathology on the performance of each of the index tests, we subsequently constructed a best-case and worst-case scenario. In the best-case scenario, index test results for all thyroid nodules with unavailable histopathology were set true: true positive or true negative. In the worst-case scenario, these were all set false: false positive or false negative. Pooled sensitivities, specificities and negative and positive LRs were estimated under these two scenarios. Together these two scenarios compose the range for the index tests’ true performance, had histopathological confirmation been available in all patients. We assessed these hypothetical scenarios for all diagnostic procedures under investigation except ultrasound. Most of the ultrasound studies retrospectively selected only thyroid nodules with available histopathology, resulting in a dissembled completeness of data. Results Literature search and study selection Our systematic literature search identified 21,355 references (Figure 1). After removing duplicates and screening titles and abstracts, 750 references were considered relevant. After full-text screening, we finally selected 165 unique articles that met our inclusion and exclusion criteria, including one article that was identified through screening of references [60]. Upon data extraction, one study was excluded from meta-analysis because it reported no benign histopathological reference tests [350]. Eighteen articles investigated two index tests [57, 61, 68, 109, 119, 248, 251, 253, 255, 257, 259, 260, 278, 306, 308, 309, 351]; one article investigated four [58]. The selected articles included 75 studies on various mutation analysis, 6 on microRNA, 18 on ICC, 56 on US, 14 on USE, 0 on CT, 3 on [99mTc] Tc-MIBI, 11 on [18F]FDG-PET, and 1 on DW-MRI.
RkJQdWJsaXNoZXIy MTk4NDMw