2084 - Clinical Context Improves Interpretation of ¹8F-Fluciclovine PET and Perfusion MRI for Distinguishing Radiation Necrosis from Tumor Recurrence Following Radiation Therapy for Brain Metastases
Presenter(s)
M. R. Harrell1, T. Kutuk1, D. Handley2, A. E. Crum1, K. Dibs1, J. C. Grecula1, R. Raval1, S. Zhu1, F. Fekrmandi1, S. Beyer1, D. M. Blakaj1, A. Chakravarti1, R. Singh3, and J. D. Palmer1; 1Department of Radiation Oncology, James Cancer Hospital/Wexner Medical Center, The Ohio State University, Columbus, OH, 2Center for Biostatistics, Department of Biomedical Informatics, The Ohio State University Wexner Medical Center, Columbus, OH, 3Eugene M. & Christine E. Lynn Cancer Institute, Baptist Health, Boca Raton, FL
Purpose/Objective(s):
Distinguishing radiation necrosis (RN) from tumor recurrence following radiation therapy (RT) for brain metastases remains challenging. Although perfusion MRI and amino-acid PET are increasingly incorporated into practice, their real-world diagnostic performance relative to final pathology remains uncertain. We hypothesized that clinical context may influence the diagnostic accuracy of PET and MRI.Materials/Methods:
We performed a retrospective review of patients with brain metastases after initial management (RT or surgery+RT) who underwent ¹8F-fluciclovine PET followed by surgical resection or biopsy due to concern for recurrent tumor or RN between 2023 and 2026. The primary outcome was pathologic diagnosis (active tumor vs RN). Collected variables included PET metrics (SUVmax, SUVmean, SUVpeak, SUVpeak-to-brain ratio, and SUVmax-to-blood pool ratio), MRI perfusion parameters (relative cerebral blood volume [rCBV], dichotomized as high vs low), interval from prior RT, and initial treatment context (RT alone vs surgery+RT). Wilcoxon rank-sum tests assessed associations between pathology and PET metrics, while Fisher’s exact tests evaluated associations between pathology and rCBV, treatment context, and interval from prior RT.
Results:
Thirty-eight lesions had pathologic confirmation (21 tumor, 17 RN). Lesions initially treated with RT alone were more often active tumor (19/26), and lesions initially treated with surgery+RT were more often RN (10/12) (p=0.002). Lesions assessed 615 days or less following initial management predominantly represented tumor (21/24), and those assessed > 615 days after initial management all represented RN (14/14) (p<0.001). On MRI, high-perfusion lesions were more often tumor (19/26) and most low-perfusion lesions were RN (10/12; p=0.002). Regarding PET, neither SUVmax (p=0.618), SUVpeak (p=0.472), or SUVpeak-to-brain ratio (p=0.159) were associated with pathology. SUVmean demonstrated the strongest association with pathology (median 1.61, p=0.044); a cutoff =1.59 correctly identified 16/21 tumors and 12/17 RN. Among 9 lesions misclassified by rCBV, SUVmean correctly identified 7 lesions (2/2 tumors and 5/7 RN). Diagnostic performance was strongest in lesions treated with RT alone (p=0.001).Conclusion:
Initial treatment management and time elapsed from initial RT prior to radiographic assessment were more strongly associated with final pathology than imaging metrics alone. Lesions evaluated shortly after initial management and those treated with RT alone were more likely to represent tumor recurrence (vs. surgery+RT or a longer time elapsed from initial management favoring RN). MRI perfusion provided useful but imperfect discrimination, and ¹8F-fluciclovine PET SUVmean may provide complementary value. These findings support integrating clinical context with employment of multiparametric imaging when evaluating post-RT lesions and warrant validation in larger cohorts.