James Cairns, Salman Arain, Amarnath Challapalli, Amalia Stavropoulos, Tamir Ali, Daniel Ward, John Sage, Georgia Ionescu, Julien M Y Willaime, Russell Frood, Amit Bahl, George Petrides, Andrew Scarsbrook
BUMP had good predictive ability for classifying bone metastases but underperformed compared with human-led PSMA-RADS assessment. For equivocal lesions BUMP offered limited added value, highlighting the importance of expert interpretation.
PURPOSE: Prostate-specific membrane antigen (PSMA) PET/CT is widely used in prostate cancer management, but unspecific bone uptake (UBU) remains a diagnostic challenge particularly with [18F]PSMA-1007. Bone uptake metastatic probability (BUMP) score - a composite imaging-clinical model, was compared with PSMA Reporting and Data System (PSMA-RADS, Version 2) for predicting skeletal metastases in UBU.
METHODS: A real-world cohort of [18F]PSMA-1007 PET/CT scans performed at three tertiary cancer centres was retrospectively analysed. Bone lesions with focal PSMA uptake were evaluated. SUVmax, CT-derived mean Hounsfield unit and androgen deprivation therapy status were recorded to calculate BUMP score for each lesion. Lesional PSMA-RADS scores were assigned by experienced radiologists. Ground truth was established using follow-up imaging, prostate-specific antigen (PSA) kinetics and multidisciplinary consensus. Receiver operating characteristic (ROC) analyses assessed diagnostic performance.
RESULTS: 130 patients (median age 70 years) with 412 bone lesions were analysed; 60.9% (n = 251) of lesions were metastatic. Median SUVmax was higher in malignant than benign lesions: 7.54 (Interquartile range (IQR) 10.3) versus 3.53 (IQR 1.51). BUMP achieved an AUROC of 0.812 (95% CI: 0.636-0.931), the optimal threshold was 0.134 yielding an F1 score of 0.800. PSMA-RADS v2.0 outperformed BUMP with higher sensitivity and specificity (98% and 96%, versus 74% and 80%). In equivocal PSMA-RADS Score 3 lesions (n = 44), BUMP showed reduced discrimination (AUROC 0.61, 95% CI 0.457-0.781).
CONCLUSION: BUMP had good predictive ability for classifying bone metastases but underperformed compared with human-led PSMA-RADS assessment. For equivocal lesions BUMP offered limited added value, highlighting the importance of expert interpretation.