Plasma cell-free DNA (cfDNA) profiling of copy number variation (CNV) to identify poor prognostic biomarkers in metastatic castration-resistant prostate cancer (mCRPC).

Background: Multiple tumour tissue studies have demonstrated the prognostic utility of CNVs in mCRPC. However, accurate assessment of CNVs in plasma cfDNA remains challenging, and prognostic significance has not been well characterized. Using a large customized panel, we correlated plasma CNVs with clinical outcomes in a contemporary cohort of mCRPC patients. Method(s): Deep targeted sequencing was performed using a 180-gene cfDNA panel (Predicine) in 56 patients commencing AR pathway inhibitors (enzalutamide or abiraterone; n = 34) or taxane chemotherapy (n = 22) at two Australian institutions. Kaplan-Meier estimates and Cox proportional-hazards models were used to correlate CNVs with progression-free survival (PFS) and overall survival (OS). Significant results were validated in an independent cohort (Mayo Clinic, n = 144). Result(s): Median follow-up was 19.4 months (mo; IQR 11.3-31.9). The most common CNVs in the Australian cohort are shown (Table). OS was significantly decreased in patients with PI3KCA gain (median 21.7 mo vs 6.6 mo, p < 0.0001 ), PTEN loss (24.8 mo vs 11.7 mo, p = 0.0019) and AR gain (21.7 mo vs 12.0 mo, p = 0.0083). Furthermore, all three alterations independently predicted for poor survival in multivariable analyses (MVA; Table). Findings in the independent cohort showed similar OS results in MVA: PIK3CA gain (HR 2.0, p = 0.07), PTEN loss (HR 1.7, p = 0.08) and AR gain (HR 1.7, p = 0.03). Conclusion(s): Sequencing of plasma cfDNA revealed that PTEN loss, and PIK3CA and AR gain are associated with inferior clinical outcomes in patients commencing contemporary systemic treatment. These data support therapeutic strategies co-targeting the PI3K and AR pathways in mCRPC. Covariates in MVA: circulating tumour DNA fraction > 2%, prior treatment, visceral disease, baseline pain and ECOG >2.