Composite Prediction Score to Interpret Bone Focal Uptake in Hormone-Sensitive Prostate Cancer Patients Imaged with [18F]PSMA-1007 PET/CT.

Abstract

Unspecific bone uptake (UBU) related to [¹⁸F]PSMA-1007 PET/CT imaging represents a clinical challenge. We aimed to assess whether a combination of clinical, biochemical, and imaging parameters could predict skeletal metastases in patients with [¹⁸F]PSMA-1007 bone focal uptake, aiding in result interpretation. Methods: We retrospectively analyzed [¹⁸F]PSMA-1007 PET/CT performed in hormone-sensitive prostate cancer (PCa) patients at 3 tertiary-level cancer centers. A fourth center was involved in performing an external validation. For each, a volume of interest was drawn using a threshold method to extract SUV_max, SUV_mean, PSMA tumor volume, and total lesion PSMA. The same volume of interest was applied to CT images to calculate the mean Hounsfield units (HU_mean) and maximum Hounsfield units. Clinical and laboratory data were collected from electronic medical records. A composite reference standard, including follow-up histopathology, biochemistry, and imaging data, was used to distinguish between PCa bone metastases and UBU. PET readers with less (n = 2) or more (n = 2) experience, masked to the reference standard, were asked to visually rate a subset of focal bone uptake (n = 178) as PCa metastases or not. Results: In total, 448 bone [¹⁸F]PSMA-1007 focal uptake specimens were identified in 267 PCa patients. Of the 448 uptake samples, 188 (41.9%) corresponded to PCa metastases. Ongoing androgen deprivation therapy at PET/CT (P < 0.001) with determination of SUV_max (P < 0.001) and HU_mean (P < 0.001) independently predicted bone metastases. A composite prediction score, the bone uptake metastatic probability (BUMP) score, achieving an area under the receiver-operating-characteristic curve (AUC) of 0.87, was validated through a 10-fold internal and external validation (n = 89 bone uptake, 51% metastatic; AUC, 0.92). The BUMP score's AUC was significantly higher than that of HU_mean (AUC, 0.62) and remained high among lesions with HU_mean in the first tertile (AUC, 0.80). A decision-curve analysis showed a higher net benefit with the score. Compared with the visual assessment, the BUMP score provided added value in terms of specificity in less-experienced PET readers (88% vs. 54%, P < 0.001). Conclusion: The BUMP score accurately distinguished UBU from bone metastases in PCa patients with [¹⁸F]PSMA-1007 focal bone uptake at PET imaging, offering additional value compared with the simple assessment of the osteoblastic CT correlate. Its use could help clinicians interpret imaging results, particularly those with less experience, potentially reducing the risk of patient overstaging.