Real-world analysis of primary tumor transcriptomes in patients subsequently treated with PARP inhibitors.

Abstract

e17140 Background: PARP inhibitors are used alone or in combination with hormonal therapies for patients with HRR-mutated metastatic castrate resistant prostate cancer (CRPC) but little is known about the transcriptional landscape of the primary tumor in these patients. Here, we explored the Decipher prostate cancer genomic classifier (GC) in such patients. Methods: Transcriptomic data from non-metastatic patients who underwent GC testing at biopsy (n=146,294) or from RP (n=51,488) between October 2016 to February 2024 obtained from the GRID registry (NCT02609269). Real-world clinical data (RWD) aggregated from insurance claims, pharmacy and electronic health records were compiled in the GRID RWD database (Veracyte, Inc, South San Francisco). PARP inhibitor (PARPi) use after diagnosis therapy was defined by the presence of one or more NDC-9 codes for olaparib, niraparib, rucaparib and talazoparib. Patients treated with PARPi were matched based on NCCN risk groups, tumor grade group (GG), or pathological features to GRID cohort. Adverse molecular features (AMF) including signatures for androgen receptor activity (AR-A), basal-luminal subtype (PSC), PTEN inactivation, TP53 mutations, RB1 loss, homologous recombination (HR) deficiency, immune activity (activated CD8 T- and T-reg cells) with Decipher score were examined using Mann-Whitney U, Kruskal-Wallis, and chi-squared tests. Results: 63 patients were treated with PARP inhibitor after transcriptome testing (27 with diagnostic biopsy tissue testing and 36 with testing from RP specimens). Median time from testing to receipt of PARPi was 42.5 months (95% CI 18.4-61.7). Overall, median Decipher score among PARPi recipients was 0.89 (IQR 0.68-0.98). PARPi recipients were matched to non-recipients who had transcriptomic testing based on NCCN or GG and surgical stage. Median Decipher score in the matched (n=5454) set was 0.77 (IQR 0.49-0.93, p<0.001) substantially higher than unmatched patients (biopsy 0.47, RP 0.63). Comparing additional AMF to the matched untreated cohort, PARPi-treated had higher prevalence of low AR-A (21% vs 14%), less basal immune subtype (40% vs 49%), higher HR deficiency scores (-0.13 vs -0.15, p<0.02), more PTEN inactivation (37% vs 21%, p=0.005), higher TP53 mutation (38% vs 29%, p<0.13) but not RB1 loss signature scores (p=1.0). Levels of activated CD8, but not Tregs, were slightly lower in PARPi. Conclusions: In this real-world study of patients initially presenting with non-metastatic disease who subsequently received PARP inhibitors, transcriptome analysis from the primary tumor revealed the presence of adverse molecular features when disease was still localized. Patients with higher Decipher score, p53 mutation signature and PTEN inactivation tumors at initial testing, could be considered for additional confirmatory genomic or genetic testing and potentially novel trials of PARPi.