<jats:title>Abstract</jats:title><jats:sec> <jats:title>Background</jats:title> <jats:p>In the phase 3 KEYNOTE-061 study (cutoff: 10/26/2017), pembrolizumab did not significantly prolong OS vs paclitaxel as second-line (2L) therapy in PD-L1 combined positive score (CPS) ≥ 1 gastric/GEJ cancer. We present results in CPS ≥ 1, ≥ 5, and ≥ 10 populations after two additional years of follow-up (cutoff: 10/07/2019).</jats:p> </jats:sec><jats:sec> <jats:title>Methods</jats:title> <jats:p>Patients were randomly allocated 1:1 to pembrolizumab 200 mg Q3W for ≤ 35 cycles or standard-dose paclitaxel. Primary endpoints: OS and PFS (CPS ≥ 1 population). HRs were calculated using stratified Cox proportional hazards models.</jats:p> </jats:sec><jats:sec> <jats:title>Results</jats:title> <jats:p>366/395 patients (92.7%) with CPS ≥ 1 died. Pembrolizumab demonstrated a trend toward improved OS vs paclitaxel in the CPS ≥ 1 population (HR, 0.81); 24-month OS rates: 19.9% vs 8.5%. Pembrolizumab incrementally increased the OS benefit with PD-L1 enrichment (CPS ≥ 5: HR, 0.72, 24-month rate, 24.2% vs 8.8%; CPS ≥ 10: 0.69, 24-month rate, 32.1% vs 10.9%). There was no difference in median PFS among treatment groups (CPS ≥ 1: HR, 1.25; CPS ≥ 5: 0.98; CPS ≥ 10: 0.79). ORR (pembrolizumab vs paclitaxel) was 16.3% vs 13.6% (CPS ≥ 1), 20.0% vs 14.3% (CPS ≥ 5), and 24.5% vs 9.1% (CPS ≥ 10); median DOR was 19.1 months vs 5.2, 32.7 vs 4.8, and NR vs 6.9, respectively. Fewer treatment-related AEs (TRAEs) occurred with pembrolizumab than paclitaxel (53% vs 84%).</jats:p> </jats:sec><jats:sec> <jats:title>Conclusion</jats:title> <jats:p>In this long-term analysis, 2L pembrolizumab did not significantly improve OS but was associated with higher 24-month OS rates than paclitaxel. Pembrolizumab also increased OS benefit with PD-L1 enrichment among patients with PD-L1-positive gastric/GEJ cancer and led to fewer TRAEs than paclitaxel.</jats:p> </jats:sec><jats:sec> <jats:title>Trial registration</jats:title> <jats:p>ClinicalTrials.gov, NCT02370498</jats:p> </jats:sec>