Quantitative assessment of PSMA imaging before and after ¹⁷⁷Lu-PSMA-617 treatment in a Ph I/II trial.

37 Background: We have previously reported a dose-intense single-cycle of 177Lu-PSMA-617 was effective in pretreated patients with mCRPC without requiring PSMA-positive imaging for enrollment. Prior post-hoc analyses of these data using approximate quantification of exclusively the most PSMA-positive disease sites have demonstrated associations between PSA response and PFS with pre-treatment 68Ga-PSMA11-PET signal. Greater sophistication in pre- and post-treatment evaluation of PSMA-expression in tumor and normal organs may allow for better patient selection and prediction of toxicities. Methods: A total of 50 patients were treated on a phase I/II study of fractionated-dose (D1, 15) 177Lu-617-PSMA. Quantification using artificial intelligence (AI) were used to measure pre- and post-treatment PSMA signal intensity. Scoring cutoffs with confidence intervals around scan variation were empirically established from a subset of test/re-test of subjects within 24h without intervening therapy. A variety of measurements were performed including SUVmean across all detectable tumor lesions, volume of lesions, and SUVtotal (Total + Volume), as well as select normal organs and changes after treatment. Associations with survival were tested via Cox proportional hazard models in univariate analyses and associations with adverse events (AEs) and PSA responses were via assessed via Wilcoxon rank sum tests. Results: 13 subjects were selected to complete AI-based quantification and associated survival analyses. Among these, 10 (77%) experienced any PSA decline, with 8 (62%) with PSA50 and 3 (23%)with PSA90. Median overall survival (OS) was 17.0 mos (10, NA) via Kaplan-Meier estimates. In univariate analysis, pretreatment SUVmean was associated with improved PFS (HR 0.66, 95% CI 0.49-0.90, p = 0.009) and OS (HR 0.81, 95% CI 0.65-1.00, p = 0.048). The metrics most strongly associated with PSA50 were pretreatment SUVmean (median [IQR]: 7.66 [6.52, 10.54] v 3.50 [3.02, 3.56], p = 0.019) and SUV Total (14982 [11110, 20595] v 1303 [576, 1512], p = 0.019), and change in Volume (-27 [-44, -20] v 145 [38, 154], p = 0.006) and SUVtotal (-57 [-67, -35] v 132 [9, 269], p = 0.030). Subjects with xerostomia had higher salivary gland SUVmax (pretreatment and change in after treatment). Those with pain flare had lower pretreatment SUV scores (Mean, Max, Total) in unaffected portions of bony skeleton. Conclusions: Sophisticated AI-based quantification analysis of PSMA expression on pre- and post-treatment 68Ga-PSMA11-PETs demonstrate associations with treatment efficacy (PSA response, OS), and associations between patient experience (AEs) and PSMA expression in non-tumor tissues. Expansion of this algorithm to a larger patient cohort may improve our ability to anticipate toxicity by body-wide PSMA detection and predict treatment response. Clinical trial information: NCT03042468.