Background: Efficacy of PD-1 inhibition has not been demonstrated in large-scale mCRPC trials. Pembro showed preliminary antitumor activity in PD-L1+ mCRPC in KEYNOTE-028 (n = 23). Here, we present results from cohorts 1-3 (n = 258) of the phase 2 KEYNOTE-199 study of pembro monotherapy in docetaxel-refractory mCRPC (NCT02787005). Methods: Cohorts 1 (C1) and 2 (C2) enrolled patients (pts) with RECIST-measurable PD-L1+ and PD-L1– disease, respectively. C3 enrolled pts with nonmeasurable, bone-predominant disease. All pts had ECOG PS 0-2 and received ≥1 novel endocrine therapy (eg, abiraterone, enzalutamide) and 1-2 prior chemotherapies including docetaxel. Pts received pembro 200 mg Q3W until PD or intolerable toxicity. Response was assessed Q9W in yr 1, then Q12W. Primary end point was ORR per RECIST v1.1 by central review in C1 and C2, separately and combined. Key secondary end points included DCR (CR + PR + SD) per PCWG3-modifed RECIST and safety in all 3 cohorts. Results: 131 pts enrolled in C1, 67 in C2, and 60 in C3. Median follow-up as of Oct 13, 2017, was 8.1 mo, 7.9 mo, and 11.8 mo, respectively. Antitumor activity was observed in all cohorts (Table). Across cohorts, DCR lasting ≥6 mo was 11%. In C1 and C2, 9% of pts had a ≥30% decrease in target lesions; 48% had target lesion changes between –30% and +20%. The response rate was numerically higher in pts with somatic BRCA1/2 or ATM mutations (12%). Drug-related grade 3-5 AE rates were 13% in C1, 12% in C2, and 17% in C3. Conclusions: Pembro shows antitumor activity and disease control with acceptable safety in pts with docetaxel-refractory mCRPC, regardless of PD-L1 status, in both RECIST-measurable and nonmeasurable disease. These data support further evaluation of pembro in mCRPC, including in pts wth homologous recombination defects. Clinical trial information: NCT02787005