Beyond PSA and PSMA: Defining PSA–PSMA discordance for FDG PET and genomic stratification in mCRPC.

194

Background: PSA kinetics and PSMA PET guide mCRPC response, yet PSA–PSMA discordance occurs in ~20–30% of patients. We hypothesized discordance reflects TP53/RB1-driven lineage plasticity with a PSMA-low, FDG-high phenotype. This study links discordance to survival and lineage-plastic genomic alterations. Methods: This prespecified retrospective biomarker analysis testing the hypothesis that PSA–PSMA discordance independently predicts worse OS and PFS versus concordant response. The null hypothesis assumed no hazard difference. We analyzed 1,850 mCRPC patients treated with ARPIs, taxanes, or 177Lu-PSMA-617. Eligibility required serum testosterone <50 ng/dL, prior ARPI or taxane, ECOG 0-2 and paired baseline and 8–12 week PSMA PET with PSA; patients without paired data were excluded. Discordance was defined as ≥50% PSA decline with RECIP 1.0 progression at 8-12 weeks confirmed by central review. Outcomes were correlated with ^18F-FDG PET/CT and TP53, RB1, PTEN alterations from tissue or cfDNA (n= 1,120). Models adjusted for LDH, ALP, visceral metastasis. Primary endpoints were OS and PFS. Powered for overall survival, assuming 20% discordance and median OS 16 months, 1,800 patients provided >90% power to detect HR 1.5 at two-sided α 0.05 with type I error 5% and type II error 10%. Survival analyses used the intention-to-treat population using Kaplan–Meier and adjusted Cox models. Genomic analyses included patients with tissue or cfDNA profiling (collected at progression or within 30 days of PSMA PET) using validated ≥300-gene NGS panels in CLIA laboratories before subsequent therapy. PSMA PET used 68Ga-PSMA-11 or 18F-DCFPyL and 18F-FDG PET used standard 60 ± 10 minute acquisition. Progression was defined per RECIP 1.0. Genomic enrichment used logistic regression. Discordance and genomic alterations were modeled as binary variables with hierarchical testing, OS first. Results: Median follow-up was 14.6 months. Discordance occurred in 20.4% (n=377), 82% from PSA decline with radiographic progression. Discordant patients had worse OS (7.2 vs 16.4 months; aHR 2.45; 95% CI 1.8–3.3; p<0.001) and PFS (3.8 vs 8.4 months; HR 2.10; 95% CI 1.6–2.8; p<0.0001). 85% were FDG-avid. TP53/RB1/PTEN alterations were enriched 3.5-fold (48% vs. 14%; p < 0.001). Findings were consistent across ARPIs (HR 2.3), taxanes (HR 2.1), and ^177Lu-PSMA-617 (HR 2.5). Features aligned with treatment-emergent small cell or AR-indifferent biology. 42% met criteria for platinum intensification. Conclusions: PSA–PSMA discordance is a clinically actionable marker of lineage plasticity and TP53/RB1-deficient mCRPC. These data support integrating 18F-FDG PET and genomic profiling to guide therapy beyond AR or PSMA targeting.