Enrico Martin

128 Chapter 6 semiquantitative parameters that adequately detect MPNSTs as well as exclude benign neurofibromas. SUVmax is the most commonly used characteristic, but ideal thresholds vary across studies. The threshold of ≥3.5 has been proposed most commonly as the ideal threshold. 15,51,57 This has been debated as several authors claim the threshold should be higher for it to be useful. Nonetheless, the threshold of ≥3.5 yielded highest accuracies across 11 different populations, which strengthens the belief that this threshold should be used. Indeed the characteristic remains imperfect as it is only has a moderately good positive likelihood ratio (4.7), meaning biopsies still play an important role as neurofibromas may also exhibit SUVmax values of ≥3.5 in 34.6% of patients in this meta-analyses. Nevertheless, the remaining 65.4% with SUVmax values of <3.5 do not require biopsies if they do not present ill-defined margins or perilesional edema on MRI. Delayed scans have been proposed to increase the accuracy of detecting MPNSTs, but it has not yet repeatedly been proven. 45,51,56 Besides, this method requires more resources and exposes patients to additional radiation. SUV measurement may additionally vary across scanners due to differing reasons. The use of proportional SUV values of tumor to tissue may be more reproducible as it reduces measurement variations. Most commonly the T/L ratio is used, but ideal thresholds are still missing. The T/L ratio did provide equal diagnostic accuracy compared to SUVmax. To diminish variations across scanners and increase reproducibility of thresholds, the European Association of Nuclear Medicine Research Ltd (EARL) set up criteria to which scanners should adhere. 66 To our knowledge, none of the studies in this review reported on a population scanned with a PET scanner that adheres to these criteria. Qualitative assessment of FDG-PET scans is also not subjected to variation in measurements and although interobserver agreement is good within studies, standardized criteria are currently lacking. Besides the use of FDG-PET scans to identify malignant transformation, it may also facilitate CT-guided biopsies and increase accuracies. 67 MPNSTs arising from plexiform neurofibromas can show heterogeneous degrees of malignancy within one tumor and are notorious for sampling errors, 18,19 thus PET-CT guided biopsies may be beneficial. Several studies in this review have shown that PET- MRI may adequately be used in the NF1 population and is particularly interesting in these patients as it combines the accuracy of both diagnostic modalities. 16,22 Moreover, replacing the CT with an MRI scan diminishes radiation exposure, which may accumulate due to numerous follow-up scans necessary in NF1 populations. 68,69 Strengths, limitations, and future perspectives Limitations to this study include the relatively high proportion of studies included to be at high risk of bias, most commonly due to concerns regarding patient selection. There was heterogeneity among study populations which led to heterogeneity of diagnostic accuracy as evaluated by subgroup analyses. Studies could be too strict in patient selection when only histologically proven lesions are included, possibly representing a group of lesions that are considered high-risk of malignancy based on imaging. Contrarily, when non-symptomatic lesions are included the proportion of low-risk