288 chapter 3 a NPV similar to a benign cytologic diagnosis (i.e., 96.3%) as per the ATA recommendations for a useful rule-out test [17]. De Koster et al. performed quantitative analysis and ROC curve analysis of the EfFECTS dataset, including 123 patients who underwent [18F]FDG PET/CT according to the EANM guidelines [455]. Quantitative [18F]FDG PET/CT assessment ruled out malignancy in indeterminate thyroid nodules, optimising the rule-out ability when distinctive SUV cut-offs were applied to Hürthle and non-Hürthle cell nodules. In non-Hürthle cell nodules, malignancy could be ruled out at a SUVmax cut-off of 2.1 g/mL (similar to visual interpretation) with a sensitivity of 96% and benign call rate of 18%. In Hürthle cell nodules, a higher cut-off at 5.2 g/mL could rule out malignancy with a sensitivity of 100% and benign call rate of 17%. As such, quantitative analysis appears advantageous over visual analysis in Hürthle cell nodules. Consequently, [18F]FDG PET/CT may be a reliable rule-out test for both non-Hürthle and Hürthle cell nodules, although external validation of these SUV thresholds is required before implementation in clinical practice. Two recent publications investigated [18F]FDG PET/CT radiomics in cytologically indeterminate thyroid nodules for the classification of malignancies [455, 459]. Giovanella et al. published the first retrospective study in 78 Bethesda III/IV patients (65 non-Hürthle nodules), suggesting a multiparametric model including cytological classification and two radiomic features [459]. The included features were the autocorrelation of the grey level cooccurrence matrix, a feature that describes the fineness of a texture, and the sphericity of the nodule shape, indicating a taller than wide shape. The cross-validated models with the two radiomic features resulted in AUCs of 73% and 73% for all nodules and in a subgroup of non-Hürthle cell nodules, respectively. In non-Hürthle cell nodules, a model with both the radiomic features and the cytological classification resulted in an AUC of 82%. A secondary analysis of the EfFECTS dataset performed additional radiomic analysis in [18F]FDG-positive scans only [455]. The authors found that radiomic analysis did not contribute to the additional differentiation of [18F]FDG-positive nodules. Both studies concluded that radiomic analysis alone on [18F]FDG PET/CT seems of no added value in the management of indeterminate thyroid nodules. However, implemented in the multiparametric model of two radiomic features and the cytological classification that Giovanella et al. proposed, clinical application of radiomics seems feasible, although validation is required. The availability of PET/CT scanners and tracers is increasing but varies worldwide. Transport distances are limited due to the short half-life of 18F (~110 minutes), which is produced in cyclotrons. The radiation exposure of an [18F]FDG PET/CT scan is mainly accounted for by the [18F]FDG dosage, which amounts about 3.5 millisievert for an administered activity of 185 MBq [363]. The radiation exposure of CT largely varies, but can be less than 0.5 millisievert for a low-dose CT of the neck region only. Costs for an investigation are generally higher than for the other modalities described, because of the costs of PET hardware and the production and transportation of radiopharmaceuticals. Two studies assessed the cost-effectiveness of an [18F]FDG PET/CT-driven management as compared to diagnostic surgery in all Bethesda III/IV patients. A 2014 cost-effectiveness model by Vriens et al.
RkJQdWJsaXNoZXIy MTk4NDMw