Cabozantinib (C) ± atezolizumab (A) in patients (pts) with metastatic castration-resistant prostate cancer (mCRPC): Expansion cohorts from the open-label phase 1b COSMIC-021 study.

145 Background: The phase 3 CONTACT-02 study significantly improved PFS and showed a trend in OS benefit with C+A vs a second novel hormone therapy (NHT) in mCRPC pts with soft tissue metastases (mets). We report outcomes from similar pts in the randomized mCRPC expansion cohorts of COSMIC-021 evaluating the contribution of adding A to C vs C alone. Methods: Pts with mCRPC with extrapelvic visceral or nodal mets who progressed on one prior NHT were randomized to receive C (60 mg QD), A (1200 mg IV Q3W), or C (40 mg QD) + A (1200 mg IV Q3W). Cohort A (n=10) was terminated early (lacked objective responses). The primary and secondary endpoints were investigator-assessed ORR per RECIST v1.1 and safety, respectively. Efficacy outcomes by blinded independent radiology committee (BIRC) were also assessed. Flow cytometry and proteomic profiling were performed on baseline and on-treatment samples. Results: A total of 101 pts were randomized to C (n=51) and C+A (n=50); median age was 70 y in both groups, 75%/72% had bone mets, 20%/24% had liver mets, and 31%/24% received docetaxel for mCSPC. ORR by BIRC (Table) and investigator (C: 20%, 95% CI, 10–33; C+A: 22%, 95% CI, 12–36) were similar between groups. Although PFS by BIRC was similar (Table, HR, 0.96; 95% CI, 0.59–1.58), the KM estimate of pts without events at 12 mo was higher with C+A (25.8% vs 14.0%). Compared with C, C+A group had longer DOR (~2-fold), fewer pts with progressive disease (PD) as best response, a numerically higher median OS (HR, 0.91; 95% CI, 0.57–1.46), more pts with PSA response with longer duration of response (median mo [95% CI], NE [6.24, NE] vs 6.93 [4.63, NE]), and prolonged time to PSA progression (Table). In the liver mets subgroup, median PFS and OS were longer with C+A vs C (PFS: HR, 0.44; 95% CI, 0.15–1.34, 5.7 vs 2.5 mo; OS: HR, 0.25; 95% CI, 0.08–0.77, 14.7 vs 5.0 mo). Grade 3–4 treatment-related adverse events occurred in 51% and 58% in C and C+A, respectively; no grade 5 events occurred. Differential regulation of activated CD8 T cells, myeloid-derived suppressor cells, and immune response and apoptosis-related proteins were observed with C+A vs C or A. Conclusions: C and C+A demonstrated clinical activity in pts with mCRPC and extrapelvic soft tissue mets, with no new safety signals. These exploratory data from randomized cohorts suggest an additive effect of A to C with a prolonged DOR in C+A vs C; additive clinical benefits were noted in pts with liver mets. Clinical trial information: NCT03170960 . Efficacy outcomes by BIRC. C (n=51) C+A (n=50) ORR, % (95% CI) 12 (4–24) 14 (6–27) Complete / partial response, n (%) 0 / 6 (12) 0 / 7 (14) Stable disease / PD, n (%) 33 (65) / 9 (18) 31 (62) / 6 (12) DCR, n (%) 39 (76) 38 (76) Median DOR (95% CI), mo 4.3 (2.8–NE) 10.6 (6.7–NE) PSA response, n (%) 5 (12%) 8 (20%) Median time to PSA progression (95% CI), mo 3.5 (1.7–5.7) 5.6 (2.7–NE) Median PFS, mo 6.7 5.5 Median OS, mo 13.8 15.6 NE, not evaluable for KM estimates.