Enrico Martin

115 Diagnostic accuracy of non-invasive tests accuracies of MRI, FDG-PET, and liquid biopsies by means of a systematic review and meta-analyses. These findings may result in characterization of lesions that obviate the need for biopsies. Methods Literature search A systematic search was performed in both PubMed and Embase databases according to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analysis) guidelines, in order to identify all potentially relevant articles between January 2000 and November 2019. The search string was built with the help of a professional librarian using search terms related to “MRI”, “PET”, “liquid biopsy”, and “MPNST”. The exact search syntaxes for PubMed and Embase are shown in Supplementary Table 1 . Studies were included when both extracranial MPNSTs and BPNSTs were evaluated and described their differences using MRI, FDG-PET, and/or liquid biopsy. Exclusion criteria were lack of full text, case reports, conference abstracts, and reviews. The initial review was conducted by two independent authors (E.M. and R.G.). Disagreements were solved through discussion, in which two additional authors were involved (D.H. and L.G.). By cross-referencing included articles, additional studies not initially included in our search were added. Data extraction Study, patient, and diagnostic test characteristics were extracted from included studies by two independent authors (R.G. and E.M.). Values of true positives (TP), false negatives (FN), false positives (FP), and true negatives (TN) were extracted per study for all mentioned diagnostic tests. Whenever this was not directly available, the rate of MPNST and provided sensitivity and specificity were used to recalculate TP, FN, FP, and TN. A true positive corresponded to an MPNST, a true negative to a BPNST. A true negative was concluded whenever the lesion was resected, there had been a biopsy with adequate follow-up or in NF1 patients the lesion was suspected to be benign and there had been adequate follow-up to exclude potential malignant transformation. Additionally, individual SUVmax values were collected when available. When the available data was insufficient for recalculation or individual SUVmax were missing, the corresponding authors of the study were requested for additional data. A reminder email was sent up to three times. Statistical analysis Using TP, FN, FP, and TN, sensitivity, specificity, and likelihood ratios were calculated for all available diagnostic tests. Sensitivity and specificity were plotted in forest plots with 95% credibility intervals (95% CI). Accuracy was based on determining the presence of an MPNST. Bayesian bivariate meta-analyses were performed on imaging 6