Dual FDG/PSMA PET imaging to predict lesion-based progression of mCRPC during PSMA-RLT

Abstract Candidates for prostate-specific membrane antigen (PSMA)-targeted radioligand therapy (RLT) of metastatic castration-resistant prostate cancer (mCRPC) frequently have “mismatch” lesions with pronounced 18-fluorodeoxyglucose ([ 18 F]FDG) but attenuated PSMA ligand uptake on positron emission tomography (PET). However, no quantitative criteria yet exist to identify mismatch lesions and predict their response to RLT. To define such criteria, we retrospectively analyzed 267 randomly-selected glucometabolic mCRPC metastases from 22 patients. On baseline PET, we determined [ 18 F]FDG and [ 68 Ga]Ga-PSMA-11 maximum standardized uptake value (SUV max ), and calculated the [ 18 F]FDG SUV max /[ 68 Ga]Ga-PSMA-11 SUV max quotient (FPQ). From follow-up [ 18 F]FDG PET after two lutetium-177-PSMA-617 RLT cycles, we evaluated the treatment response and categorized the lesions into three subgroups (partial remission, stable disease, progression) based on change in [ 18 F]FDG SUV max . Lastly, we compared the baseline PET variables in progressing versus non-progressing lesions. Variables differing significantly, and a score incorporating them, were assessed via receiver operator characteristic (ROC) curve analysis, regarding ability to predict lesional progression, with area under the curve (AUC) as metric. Cut-offs with optimal sensitivity and specificity were determined using the maximum value of Youden's index. Fifty-one of 267 lesions (19.1%) progressed, 102/267 (38.2%) manifested stable disease, and 114/267 (42.7%) partially responded after two RLT cycles. At baseline, median [ 68 Ga]Ga-PSMA-11 SUV max was significantly lower ( p < 0.001), median FPQ significantly higher ( p < 0.001), and median [ 18 F]FDG SUV max similar in progressing versus non-progressing lesions. [ 68 Ga]Ga-PSMA-11 SUV max and FPQ showed predictive power regarding progression (AUCs: 0.89, 0.90). An introduced clinical score combining both further improved predictive performance (AUC: 0.94). Optimal cut-offs to foretell progression were: [ 68 Ga]Ga-PSMA-11 SUV max < 11.09 (88.2% sensitivity, 81.9% specificity), FPQ ≥ 0.92 (90.2% sensitivity, 78.7% specificity), clinical score ≥ 6/9 points (88.2% sensitivity, 87.5% specificity). At baseline, a low [ 68 Ga]Ga-PSMA-11 SUV max and a high FPQ predict early lesional progression under RLT; [ 18 F]FDG SUV max does not. A score combining [ 68 Ga]Ga-PSMA-11 SUV max and FPQ predicts early lesional progression even more effectively and might therefore be useful to quantitatively identify mismatch lesions.