Abstract
394
Objectives: Gliomas are the most common primary brain tumor in the US and account for approximately 25% of all brain tumors, and 80% of all malignant brain tumors.1 The amino acid positron emission tomography (PET) tracer, 18F-fluciclovine, has Orphan Drug designation in the US and Europe for imaging glioma and has the potential to address the limited utility of the current standard of care, magnetic resonance imaging (MRI).2 However, in order to facilitate the successful migration of this technology from expert sites into general clinical use an understanding of the reproducibility of interpretation of 18F-fluciclovine images by naïve users is required. Here, we explore 3D Volumes of Interest (VOI) derived from 18F-fluciclovine imaging of suspected glioma and report the reproducibility of 18F-fluciclovine PET image interpretation among naïve users. Data were captured as part of a Phase 3, blinded image evaluation of the diagnostic performance and reproducibility of interpretation of 18F-fluciclovine images when interpreted in combination with contrast-enhanced T1-weighted (CE-T1W) MRI for suspected glioma (abbreviated as 18F-fluciclovine + CE-T1W).
Methods: Thirty-five 18F-fluciclovine PET and MRI (CE-T1W and fluid-attenuated inversion recovery [FLAIR] [or T2W]) datasets with corresponding histopathological truth standards collected as part of a previous prospective Phase 2 trial (JapicCTI-132289) were evaluated. A neuroradiologist read the MRI images and 3 blinded nuclear medicine physicians who were naïve to 18F-fluciclovine PET independently read each 18F-fluciclovine image in combination with CE-T1W MRI. Diagnostic performance parameters were determined for 18F-fluciclovine + CE-T1W and compared with those determined for CE-T1W and FLAIR (or T2W) alone. Readers delineated the extent of any malignancy, recording the VOI. Inter- and intra-reader reproducibility were assessed for both the diagnostic performance parameters and the VOI (using intersections and similarity metrics such as dice coefficient [DC] and concordance index [CI]).
Results: The positive predictive value for 18F-fluciclovine PET + CE-T1W MRI was > 90% for all 3 readers, similar to CE-T1W MRI alone (94%), but higher than that of FLAIR (85%). The sensitivity of 18F-fluciclovine PET + CE-T1W MRI (66-71%) was higher than that of CE-T1W MRI alone (42%) and the specificity (89%) was higher than that of FLAIR (or T2W) MRI alone (33%). Inter‑reader concordance based on the binary interpretation of the data between the 3 readers for 18F-fluciclovine PET + CE‑T1W was 89% (Fleiss’ Kappa, 0.86). Inter‑reader VOI agreement was also high (median DC, 0.78−0.89; median CI, 0.64−0.80 between comparisons). Intra‑reader concordance was 83.3−100% (Cohen’s Kappa, 0.66−1.00) for the three readers interpreting an image for the second time, with the DC for second VOI recordings ranging from 0.90 to 0.91 and the median CI ranging from 0.81 to 0.84 across the 3 readers. When compared with the VOI identified on CE-T1W MRI alone, a median increase in VOI of 28−60% was recorded with 18F-Fluciclovine PET + CE-T1W MRI. Comparisons with FLAIR imaging showed a median additional volume identified on FLAIR (or T2W) MRI, but not on 18F-fluciclovine PET + CE-T1W MRI, of 67−68% of the total volume identified by FLAIR (or T2W) MRI.
Conclusions: 18F-Fluciclovine PET + CE-T1W MRI shows good diagnostic performance, with consistent image interpretation demonstrated across 3 naïve readers. 18F-Fluciclovine PET + CE-T1W MRI accurately identified higher tumor volumes than CE-T1W MRI alone and showed greater specificity than FLAIR (or T2W) MRI. The addition of 18F-fluciclovine PET to MRI has potential clinical value in delineating which part of the additional volume of abnormality shown on FLAIR is not malignant, thus more accurately identifying the extent of glioma prior to surgery or radiotherapy.