Background: Pembrolizumab provides effective antitumor immunity and durable responses in patients (pts) with advanced, colorectal cancer (CRC) with microsatellite instability-high (MSI-H) tumors. We present data on antitumor immunity with pembrolizumab in pts from the phase 2, KEYNOTE-164 study who had approximately 3 years of follow-up, and in pts re-treated after disease progression. Methods: KEYNOTE-164 enrolled pts with metastatic MSI-H CRC, MSI-H status confirmed locally by IHC or PCR, and ≥2 (cohort A) or ≥1 (cohort B) prior lines of therapy (fluoropyrimidine, oxaliplatin, irinotecan, or anti VEGF/EGFR). Eligible pts received pembrolizumab 200 mg Q3W for 2y (35 administrations) or until progression, unacceptable toxicity, or withdrawal. Pts who stopped pembro due to a confirmed CR or after completing 2y of treatment and who progressed after stopping were eligible for re-treatment with up to 17 administrations in the second-course phase, at investigator discretion. Tumor response was assessed Q9W per RECIST v1.1 by independent review. The primary endpoint was ORR. Secondary endpoints included DOR, PFS, OS, and safety. The data cutoff date was Sep 9, 2019. Results: At data cutoff, the median follow-up was 31.4 mo (range, 0.2-47.8) for 61 pts in cohort A and 36.1 mo (0.1-39.3) for 63 pts in cohort B. ORR was 32.8% (3CR, 17PR; 95% CI% 21.3-46.0) for cohort A and 34.9% (8CR, 14PR; 95% CI 23.3-48.0) in cohort B. Median DOR was not reached (NR [range, 6.2-41.3+]) and not reached (range, 3.9+ to 37.1+), respectively. Fifteen pts in cohort A and 17 in cohort B had ongoing responses at data cutoff. Median PFS was 2.3 mo (95% CI 2.1-8.1) with 3-yr PFS rate of 31% in cohort A and was 4.1 mo (2.1-18.9) with 3-yr PFS rate of 34% in cohort B. Median OS was 31.4 mo (21.4-NR) with 3-yr OS rate of 49% in cohort A and was not reached (19.2-NR) with 3-yr OS rate of 52% in cohort B. Nine pts (6 in cohort A, 3 in cohort B) had a second course of treatment. The best response in second course was PR in 1 patient each in cohort A and B. Grade 3-4 drug-related adverse events occurred in 10 (16%) pts in cohort A and 8 (13%) pts in cohort B. No grade 5 drug-related events occurred. Conclusions: After approximately 3 y of follow-up, pembrolizumab continues to provide effective long-term antitumor immunity with durable responses, with small numbers of drug-related adverse events and no drug-related deaths in pts with advanced, MSI-H CRC. Clinical trial information: NCT02460198.