Background: The prognosis of refractory mCRPC remains poor despite advancements in therapeutic options. KeyNote-199 demonstrated modest activity of Pem in mCRPC with expected safety profile. We present our real-world experience with Pem in mCRPC. Methods: We conducted a retrospective review of mCRPC patients treated with Pem at our institution from 1/1/2017 to 10/1/22. Baseline demographic, clinicopathologic, and genomic characteristics were recorded. PSA and radiographic responses were assessed by the study team, and survival distributions estimated using the Kaplan-Meier method. Results: A total of 39 patients were identified – 97% (37) were White, median age was 71 years, 4% (19/35) had a Gleason Score ≥8; 80% (31) had skeletal and 74% (29) had soft tissue metastases at Pem initiation. Overall, patients were heavily pre-treated (median of 7 prior therapies, range 0-8) - 87% (34) had received taxanes, 82% (32) novel antiandrogens, 23% (9) Ra-223, 21% (8) Sipuleucel-T, and 2% (1) Olaparib. Median duration on Pem was 7 months (range = 1-29). Among the 34 evaluable patients, 2 (6%) achieved CR, 2 (6%) had PR, 5 (15%) had stable disease (SD), and 25 (73%) had progressive disease (PD) on radiographic assessment. PSA reduction ≥ 50% was noted in 7/32 (22%) patients. The 4 patients who had radiographic CR/PR had positive predictive biomarkers – Patient 1: CR – MSI-H, high TMB (17.5/Mb); Patient 2: CR – MSI-indeterminate, germline MSH6 mutation; Patient 3: PR – MSI-H, high TMB (28.8/Mb), germline MSH2 mutation; and Patient 4: PR – MSI-S, high TMB (18.3/Mb), PDL1 TPS 100%, positive neuroendocrine markers. Interestingly, patient 3 was switched to ipilimumab + nivolumab after PD on Pem, and subsequently had a CR. None of the evaluated patients with SD or PD had high MSI, TMB, or PDL1 levels. The median overall survival from Pem initiation was 4.4 months (95% CI 3.0-10.2 months). Three (8%) patients discontinued Pem due to immune-related adverse effects (IRAEs); no treatment-related deaths were reported. The most frequent Gr 3 IRAEs are shown. Conclusions: Single-agent Pem demonstrated modest overall efficacy in mCRPC, restricted only to patients with predictive biomarkers. Given the non-trivial risk of IRAEs, financial toxicity, and potential QoL implications, we suggest using checkpoint inhibitors only in appropriately biomarker-selected patients with mCRPC.