390 Background: PAM50 captures prostate cancer lineage (Luminal B vs Non-Luminal B Luminal A + Basal) and may inform response to androgen-pathway therapy, while Decipher (GC) guides selection of ADT with salvage radiotherapy (sRT). We examined how PAM50 tracks with GC and whether ADT’s association with metastasis varies by subtype. Methods: Single-institution, prospectively collected post-RP cohort with biochemical recurrence. RP tissue underwent PAM50 and Decipher testing. Primary endpoint was metastasis; death before metastasis was a competing risk. We fit a Fine–Gray model with subtype (Lum B vs Non-Lum B), ADT with first-course sRT (yes/no), and the subtype×ADT interaction. Time origin was radical prostatectomy. Among 122 analyzable men, ADT distribution at time 0 was: Lum B/no ADT = 12, Lum B/ADT = 40, Non-Lum B/no ADT = 13, Non-Lum B/ADT = 57. Results: N=123 overall (LumA 30, LumB 53, Basal 40). Decipher differed by subtype: median GC LumA 0.60 (IQR 0.30–0.80), LumB 0.80 (0.60–0.90), Basal 0.70 (0.40–0.80); p=0.004. Decipher Risk groups (.84) also differed (p=0.006); GC 0.85–1.00 occurred in 41.5% of LumB vs 22.5% Basal and 13.3% LumA. Many Non-Lum B crossed thresholds commonly used to support ADT with sRT: GC > 0.45 ~63%, GC ≥ 0.60 ~53%, GC ≥ 0.85 ~19% (weighted from LumA+Basal). Basal had more positive margins (65.0% vs 43.4% LumB vs 23.3% LumA; p=0.026). In the competing-risks model (19 metastasis, 3 competing deaths), ADT was associated with lower metastasis hazard in Lum B (sHR 0.10, 95% CI 0.038–0.266; p<0.001; reference = Lum B/no ADT). Non-Lum B vs Lum B (no ADT) trended toward lower hazard (sHR 0.23, 95% CI 0.044–1.206; p=0.082). The interaction (Non-Lum B×ADT) was not significant (sHR 2.49; p=0.364), providing no evidence that the ADT association differed by subtype. Descriptively, the highest metastasis incidence occurred in Lum B/no ADT. Conclusions: PAM50 correlates with genomic risk (Lum B highest GC), yet many Non-Lum B tumors also meet GC cutpoints often used to support ADT with sRT. In this cohort ADT was strongly associated with lower metastasis, and we did not detect a subtype-specific interaction (limited by 19 events and observational treatment assignment). Clinical implication: Do not withhold ADT solely because a tumor is Non-Lum B—particularly when Decipher is high—while Lum B appears to derive the greatest benefit. Basal’s higher margin positivity and longer RP→sRT interval suggest a locoregional-failure phenotype, supporting evaluation of earlier and/or locally intensified salvage.
Jonathan D. Tward (Sun,) studied this question.