Quantitative Thresholds on ¹⁸F-DCFPyL PET for Determining Malignancy in Prostate Cancer with Pathologic Correlation.

Abstract

In this article, we assess histologically validated quantitative parameters on ¹⁸F-piflufolastat (¹⁸F-DCFPyL) prostate-specific membrane antigen (PSMA) PET and identify thresholds for differentiating malignant prostate cancer from benign tissue. <b>Methods:</b> In this multicenter retrospective study, men with prostate cancer who underwent ¹⁸F-DCFPyL PET/CT followed by surgery or biopsy at 4 institutions were included. Histopathology reports and CT-guided biopsy images were anatomically matched to PET/CT, and corresponding regions of interest were delineated. Univariable associations between lesion malignancy and PET parameters (SUV_max/mean, total lesion uptake (TLU), tumor volume (TV), ratios to blood, liver, and spleen) were tested using likelihood ratio tests. Multivariable models incorporating PET and clinical variables were also evaluated. Analyses were performed for all lesions combined and separately for prostate and metastatic lesions. <b>Results:</b> In total, 305 men with 476 PET regions of interest matched to histologic specimens were analyzed. Across all lesions, malignancy was associated with all PSMA PET parameters (<i>P</i> = 0.011 for TV; <i>P</i> < 0.001 for all other metrics). Similar associations were observed in metastatic lesions (<i>P</i> = 0.001 for TV; <i>P</i> < 0.001 for all other metrics). In prostate lesions, malignancy was associated with all PET parameters except TV (<i>P</i> = 0.970 for TV; <i>P</i> < 0.001 for all other metrics). Among all evaluated metrics, SUV_max to blood-pool ratio showed the highest discriminatory performance across all lesions, prostate lesions, and metastatic lesions, with thresholds ranging from 2.86 to 2.97, sensitivities from 0.841 to 0.950, and specificities from 0.788 to 0.911. Ratios of TLU to blood-pool, liver, and spleen also performed well. SUV_max, SUV_mean, and TLU yielded good sensitivity across cohorts, although specificity was lower in some subgroups. For prostate lesions, the SUV_max threshold of 5.43 yielded a sensitivity of 0.843 and a specificity of 0.927. Multivariable models combining PET parameters, particularly blood-pool and liver-referenced metrics, with clinical variables further improved discrimination between malignant and benign lesions. <b>Conclusion:</b> SUV_max-to-blood and TLU-to-blood cutoffs can distinguish malignant prostate cancer from benign tissue and may support clinical decision-making, including biopsy selection. Larger studies are required for further confirmation.