Background: Blockade of the PD-L1/PD-1 axis is a proven immunologic approach for treatment of many cancers; however, not all pts respond to monotherapy. Bev, an anti–VEGF-A antibody, has demonstrated clinical efficacy in mCRC and enhanced T-cell infiltration in tumors in preclinical studies. Thus, we postulated that combining atezo (anti-PD-L1) with bev would augment anti-tumor immune responses, resulting in improved and more durable clinical benefit. We report results from the first study of an anti-PD-L1 agent + VEGF-A blockade in MSI-high mCRC. Methods: A Ph Ib study (NCT01633970) investigated atezo + various chemotherapeutic/biologic regimens (eg, bev) in pts with advanced solid tumors, including mCRC. Pts received atezo 1200 mg q3w plus bev 15 mg/kg q3w (data cutoff, May 20, 2016). The primary objective was to evaluate the safety of atezo + bev. Secondary objectives included anti-tumor activity per RECIST v1.1. MSI status was tested locally. Results: Ten MSI-high mCRC pts were enrolled. Three pts had received 1 prior chemotherapy and 7 pts had received ≥ 2. Median age was 52.5 y. The minimum (range) of safety follow-up was 2.6 (2.6-20.3) mo. Median (range) treatment duration with atezo + bev was 10.1 (2-20) and 9.0 (2-19) mo, respectively. Efficacy results are shown below; the confirmed ORR per RECIST v1.1 was 30% (95% CI, 6.7%-65.3%). Median OS had not been reached with a median follow-up of 11.1 mo. Treatment-related all-grade AEs occurred in 80% of pts; 40% of pts had a related G3/4 AE. The most common related AE was proteinuria (40%; n = 3 G2 and n = 1 G3). No G5 events occurred. One AE led to discontinuation of atezo and 3 AEs led to discontinuation of bev. Biomarker data will be presented. Conclusions: Initial clinical activity was observed in heavily pretreated pts with MSI-high mCRC receiving atezo + bev; the disease control rate was 90%. This combination was well tolerated without unexpected toxicities. Further follow-up is ongoing. Clinical trial information: NCT01633970

^aConfirmed responses. NE, not estimable.