‘One button push’ fully automated PSMA PET quantification: Correlation with progression free and overall survival in patients undergoing [ ¹⁷⁷ Lu] Lu PSMA therapy for metastatic castrate resistant prostate cancer.

Abstract

5054 Background: [ 177 Lu]Lu-PSMA is an effective treatment in metastatic castrate-resistant prostate cancer (mCRPC). Whole body standardized uptake value (SUV)mean and total tumor volume (PSMA-TTV) are valuable screening biomarkers for 177 Lu-PSMA therapy but require labour intensive semi-quantitative software. This study aims to compare PSMA SUVmean, and PSMA-TTV from fully automated and semi-automated methods of PSMA-PET quantification for predictive and prognostic capability. Methods: Datasets of participants (pts) from ethics approved trials with mCRPC post androgen receptor signaling inhibition and post taxane (or unfit for taxane), treated with [ 177 Lu]Lu-PSMA with a prior screening 68 Ga-PSMA-11 PET/CT, and outcome data including PSA progression-free (PSA-PFS) and overall survival (OS) were included. Screening 68 Ga-PSMA-11 PET/CT of participants were quantified using MIM LesionID Pro to derive SUVmean and PSMA-TTV with a fully automated quantification process (Method A) and semi-automated quantification adjusted manually for error (Method B). Both methods utilised software that segmented all lesions above SUVmax 3 and a CT-based deep learning method to identify normal organs for automatic physiological uptake removal. SUVmean and PSMA-TTV were evaluated in quartiles. Associations between SUVmean and PSMA-TTV above and below the 75 th percentile (Q4 vs Q1-3) were examined with Kaplan Meier estimates and log-rank tests. Results: Data from 139 pts were analysed, median age 72 years (IQR: 67–77) and median PSA 94 ng/ml (IQR: 34–325). The median time to PSA-PFS (120 events) 5.5 months (95%CI:4–6.0) and OS (82 events) 13.5 months (95%CI:11– 18). With method A (fully automated), SUVmean Q4 was 9.7 and PSMA -TTV Q4 was 1156ml. The corresponding results with method B (manually adjusted) were SUVmean Q4 9.9 and PSMA-TTV Q4 1203ml. Withmethod A, median PSA-PFS for SUVmean Q1-3 was 4.5 (95%CI:3–6) vs 7 months (mo) (95%CI:5–11) for SUVmean Q4 (p=0.003). Median OS for SUVmean Q1-3 was 12.0 (95%CI:10–6) vs 20 mo (95%CI:12.0–NE) for SUVmean Q4 (p=0.011). For PSMA-TTV Q4 vs Q1-3, median OS was 8.5 (95%CI:7 –12.0) vs 18 mo (95%CI: 13–20) (p<0.001). With method B, median PSA-PFS for SUVmean Q1-3 was 4.5 (95%CI:3– 6) vs 7.5 mo (95%CI:5–11) for SUVmean Q4 (p=0.002). Median OS for SUVmean Q1-3 was 13 (95%CI:10–17) vs 20 mo (95%CI:11 – NE) for SUVmean Q4 (p=0.03). For PSMA-TTV Q4 vs Q1-3, median OS was 8.5 (95%CI:7–12) vs 18 mo (95%CI:13 – 20) (p<0.001). Conclusions: PSMA SUVmean and PSMA-TTV with a fully automated quantification method predicted both PSA-PFS and OS in patients undergoing [ 177 Lu]Lu-PSMA therapy. Fully automated vs manually adjusted predictive capability was not different. This is an important step in moving PSMA-PET quantitative biomarkers from research tool to routine clinical care.