Using real-world outcomes to evaluate the predictive power of tissue-assessed genomic biomarkers for taxane versus novel hormonal therapy (NHT) outcomes in metastatic castration-resistant prostate cancer (mCRPC).

Abstract

5054 Background: No established genomic biomarkers exist for guiding treatment decisions between novel hormonal therapy (NHT) vs taxane chemotherapy in mCRPC. However, specific alterations in AR have been associated to decreased responsiveness to NHT in this setting. Leveraging routine comprehensive genomic profiling (CGP) testing of mCRPC tissue samples, we hypothesized that patients (pts) with AR amplification ( ARamp) would have better outcomes on taxanes over NHT. Methods: Pts were selected from Flatiron Health (FH)-Foundation Medicine (FMI) clinico-genomic database (CGDB), a nationwide deidentified electronic health record database linked to CGP. Data originated from approximately 280 US cancer clinics (̃800 sites). CGP results (including analysis of AR and 15 other genomic biomarkers) were obtained from mCRPC tumor tissue collected up to 180 days before or 30 days after initiation of new systemic therapy between 1/1/11 - 6/30/20, and linked to PSA response, time to next therapy (TTNT) and overall survival (OS). Multivariable treatment interaction models were adjusted for drug assignment imbalances (line of therapy, prior NHT or taxane use, PSA, alkaline phosphatase, hemoglobin, albumin, years to CRPC, biopsy site) using inverse probability of treatment weighting via propensity scores. Results: Among 5754 evaluable mCRPC lines of therapy, 180 receiving NHT and 179 receiving taxanes met inclusion criteria, 359 total from 308 unique patients. Pts with ARamp vs no ARamp on NHT had worse PSA response (median +57.3% vs. -31.4%, p = 0.002), TTNT (HR: 2.03, p < 0.001), and OS (HR: 2.28, p < 0.001), but had no difference in outcomes on taxanes. Multivariable interaction Cox models found ARamp independently associated to better TTNT on taxanes vs. NHT (HR: 0.48, p = 0.010), similar to pts with RB1 alterations (HR: 0.46, p = 0.027). Consistent treatment interactions were seen with OS for ARamp (HR: 0.53, p = 0.025) and RB1 (HR: 0.32, p = 0.024). While CDK12 was not predictive, it independently associated with worse OS overall (HR: 2.25, p = 0.0011). In the 55 pts who received NHT followed by taxane immediately after, ARamp pre-NHT was associated with better TTNT on the subsequent taxane than on the initial NHT (HR: 0.40, p = 0.028). Of these, 33 had PSA responses evaluable, and ARamp pre-NHT was significantly associated with better PSA decline on the subsequent taxane, despite disadvantage of first progressing on NHT (OR: 10.9, p = 0.021). Conclusions: Genomic biomarkers routinely identified with CGP such as ARamp may aid in identifying mCRPC pts who are to obtain greater benefit from taxane chemotherapy instead of NHT. Prospective efforts are needed to further validate the utility of CGP for assisting treatment decisions for mCRPC patients.