287 Non-invasive imaging biomarkers 3 with indeterminate cytology, defined as inconclusive fine-needle aspiration biopsy (pre-Bethesda), who subsequently underwent diagnostic hemithyroidectomy [37]. They demonstrated that a negative [18F]FDG PET/CT could theoretically reduce the number of futile hemithyroidectomies by 66% at a NPV of 100%. A subsequent meta-analysis from 2011 by Vriens et al., including six studies, presented a pooled sensitivity of 95% and a pooled specificity of 48%, resulting in a NPV and PPV of 96% and 39% respectively (benign call rate: 37%) [304]. In 2017, a review by De Koster et al. reported sensitivities and specificities of [18F]FDG PET/CT to detect thyroid carcinoma in indeterminate thyroid nodules ranging from 77% to 100% and 33% to 64%, respectively [25]. These findings were recently validated in a recent multicentric diagnostic randomised controlled trial that assessed the impact of [18F]FDG PET/CT in the management of thyroid nodules with double-read Bethesda III or IV cytology to rule out malignancy, avoid futile diagnostic surgeries, and improve patient outcomes (EfFECTS trial) [454]. De Koster et al. randomised 132 patients with an indeterminate nodule who were scheduled for diagnostic surgery and underwent an [18F]FDG PET/ CT scan into a PET/CT-driven arm or a diagnostic surgery arm. In the PET/CT-driven arm, diagnostic surgery was advised in visually [18F]FDG-positive nodules and active surveillance in [18F]FDG-negative nodules. In the diagnostic surgery arm, all patients were advised to continue with the scheduled diagnostic surgery. Patient management was considered unbeneficial (i.e., diagnostic surgery for benign nodules or active surveillance for malignant/borderline nodules) in 42% of patients in the [18F]FDG PET/CT-driven arm and 83% in the diagnostic surgery arm. No wrongful active surveillance for malignant/borderline nodules was reported. As such, [18F]FDG PET/CT-driven management avoided 40% of diagnostic surgeries for benign nodules. Therapeutic yield was the highest (48% reduction in diagnostic surgeries) when only non-Hürthle cell nodules were considered, as nearly all Hürthle cell nodules were [18F]FDG-positive on visual interpretation. Several studies have reported the quantitative assessment of [18F]FDG PET/CT images using the SUV of the indeterminate thyroid nodule, with a higher SUVmax reported in thyroid malignancies than in benign lesions [40, 309, 444, 455-458]. Nevertheless, major variations in SUV cut-offs and diagnostic accuracy are found between studies. Deandreis et al. and Rosario et al. respectively included 56 indeterminate nodules (pre-Bethesda) and 63 Bethesda III/IV nodules and showed that a SUVmax cut-off of at least 5 g/mL was 91% specific to detect thyroid carcinoma, NIFTP, and FT-UMP [309, 456]. This was substantiated by Piccardo et al. in 111 indeterminate nodules, but no AUCs or corresponding sensitivity and specificity were reported [40]. Contrarily, Merten et al. demonstrated that a cut-off of 5 g/mL was only 41% specific but 80% sensitive in their study in 51 Bethesda IV nodules [458]. Pathak et al. reported a SUVmax cut-off of 3.25 g/mL best differentiated 42 non-Hürthle cell nodules with 79% sensitivity and 83% specificity [457]. An additional analysis of the EfFECTS trial dataset assessed the added value of SUV metrics, SUV-ratios (node to contralateral normal thyroid) and radiomics for the preoperative differentiation [455]. None of these previous studies used ROC curve analysis to determine SUV cut-offs that corresponded to optimal test sensitivity, i.e.,
RkJQdWJsaXNoZXIy MTk4NDMw