Utility of 18F-FDG PET with a semi-quantitative index in the detection of sarcomatous transformation in patients with neurofibromatosis type 1.

BACKGROUND

Malignant peripheral nerve sheath tumors (MPNSTs) are a serious complications of neurofibromatosis type 1 associated with poor prognosis and deeper lesions can be difficult to diagnose. 18-FDG PET improves the detection of malignancies. However, the criteria for malignancy, notably the SUVmax threshold, are not standardized. Therefore, the aim of the study was to evaluate a semi-quantitative index for the reproducible detection of MPNST with FDG PET.

METHODS

It is a multicenter retrospective study conducted between 2000 to 2012. All patients with NF1 referred for suspected MPNST underwent PET. Since SUVmax was not available until 2004 in our centers, we had to settle for the semi-quantitative method used at that time, the uptake ratio between the tumor and the normal liver (T/L ratio) with 1.5 as the cut-off for malignancy. When dedicated PET with SUVmax became available, the semi-quantitative analysis of PET images remained, along with SUVmax.

RESULTS

113 patients with 145 tumors were included. PET assessment revealed 65 suspected lesions with T/L >1.5 and among these, 40 were MPNSTs. 80 tumors were classified as non-suspicious, and 79 were benign. The 1.5 T/L cut-off had a negative predictive value (NPV) of 98,8% and a positive predictive value of 61,5%. The positive likelihood ratio (LR) was 4,059, the negative LR was 0,032 with 97% sensitivity and 76% specificity.

CONCLUSIONS

This study, which is among the largest published, confirms the utility of PET for detecting NF1-associated MPNSTs. A semi-quantitative index, the T/L ratio with a cut-off of 1.5, allowed sensitive and specific differentiation of malignant from benign tumors better than SUVmax. When T/L was <1.5, MPNSTs were ruled out with 98,8% NPV. When T/L was >1.5, there was a strong suspicion of malignancy. This semi-quantitative analytical method is as simple as SUVmax, but is more sensitive, more reproducible and non-user-dependent.