Early detection and localization of recurrent prostate cancer using $$^{18}$$F-DCFPyL PET/CT in patients with low and ultra-low PSA levels

This pooled retrospective analysis included patients with prostate cancer with BCR at minimally detectable PSA levels and underwent \(^{18}\) F-DCFPyL PET/CT. Patients were identified from investigator-initiated trials at the West Los Angeles Veterans Affairs Medical Center, and the phase III CONDOR trial (NCT03739684). Local Institutional Review Board waiver for retrospective analysis on de-identified data was received at West Los Angeles Veterans Affairs Medical Center. Eligible patients had PSA values below 0.5 ng/mL at time of imaging. BCR was defined as a rising PSA after definitive therapy. Among the included patients, 95 had undergone RP, while 11 had received RT following RP; detailed information on the indication for post-prostatectomy RT was not available. Inclusion was independent of prior conventional imaging findings. Patients were stratified in two cohorts based on PSA levels at time of PET/CT: an ultra-low PSA cohort (PSA \(\le 0.2\) ng/mL) and a low PSA cohort (PSA 0.2–0.5 ng/mL). The performance of \(^{18}\) F-DCFPyL PET/CT was evaluated overall and separately within each PSA cohort. Imaging was performed according to the protocols of the contributing studies, which were closely comparable across sites. \(^{18}\) F-DCFPyL was administered intravenously at a target activity of approximately 333 MBq (9 mCi), followed by an uptake interval of 1–2 hours before acquisition. Patients voided before scanning. PET and non-contrast CT images were acquired from the skull vertex to the mid-thigh on the clinical PET/CT scanners of the contributing sites; images were reconstructed using each site’s standard iterative reconstruction protocol with CT-based attenuation correction. As imaging was pooled retrospectively from investigator-initiated trials at the West Los Angeles Veterans Affairs Medical Center and the phase III CONDOR trial, exact acquisition and reconstruction parameters varied across the pooled source datasets. Interpretation of the \(^{18}\) F-DCFPyL PET/CT was supported by the automated Prostate Cancer Molecular Imaging Standardized Evaluation (aPROMISE) platform [ 18 , 19 ]. The aPROMISE automated segmentations were independently reviewed and confirmed by an experienced, board-certified nuclear medicine physician. Lesion positivity was determined by visual assessment of focal tracer uptake above local background in an anatomically suspicious location, in accordance with established PSMA PET reading frameworks [ 20 , 21 ]. A fixed standardized uptake value (SUV) threshold was not applied. Findings consistent with physiologic PSMA uptake were excluded. The primary endpoint was detection rate, defined as the proportion of patients with at least one PSMA-avid lesion identified on PET/CT relative to the total number of patients with BCR. PET-positive findings were defined as PSMA-avid lesions interpreted as suspicious for recurrent prostate cancer on \(^{18}\) F-DCFPyL PET/CT. Since an independent reference standard was not available for all detected lesions, the analysis focused on detection rate, rather than sensitivity. Descriptive statistics were used to summarize patient characteristics and imaging findings, with categorical variables reported as counts and percentages. This retrospective analysis includes a total of 106 patients. At the time of the \(^{18}\) F-DCFPyL PET/CT, 24 patients were classified in the ultra-low PSA cohort (PSA \(\le 0.2\) ng/mL), and 82 patients in the low PSA cohort (0.2–0.5 ng/mL). Baseline patient characteristics are summarized in Table 1 . Initial International Society of Urological Pathology (ISUP) grade group ranged from 1 to 5, with the majority of patients classified as grade group 2 or 3, reflecting a predominantly intermediate-risk population. Representative examples of PSMA-avid metastatic lesions identified in ultra-low and low PSA cohorts are shown in Figs. 1 and 2 , respectively. The detection rate for the whole study population was 45%. Detection rates in the ultra-low and low PSA cohorts were 33% and 49% respectively (Table 2 ). PSMA-avid lesions were identified in 8/24 (33%) patients in the ultra-low PSA cohort and 40/82 (49%) patients in the low PSA cohort, accounting for 17 and 83 lesions, respectively. Lesion-level and region-based findings are summarized in Table 3 . In the ultra-low PSA cohort, nodal and bone lesions were observed at similar frequencies, while lymph node involvement was most frequent in the low PSA cohort. The total lesion SUV mean ranged from 2.6 to 6.0 for the PSA 0–0.2 ng/mL group and 1.8 to 7.1 for the 0.2–0.5 ng/mL group.