Background: In KEYNOTE-010 (NCT01905657), pembro was superior to doce for OS in patients (pts) with PD-L1–positive advanced NSCLC that progressed after ≥2 platinum-doublet chemotherapy cycles (P = .0002 for 2 mg/kg, P< .0001 for 10 mg/kg for PD-L1 tumor proportion score [TPS] ≥50%; P= .0008 and < .0001 for PD-L1 TPS ≥1%). We compared outcomes in pts enrolled based on PD-L1 in archival vs new tumor samples. Methods: PD-L1 was assessed centrally by IHC (22C3 antibody) in archival or new (ie, no intervening therapy between collection and start of study drug) tumor samples. 1034 pts received pembro 2 or 10 mg/kg Q3W or doce 75 mg/m²Q3W for 24 mo or until progression, intolerable toxicity, or other reason. Response was assessed by RECIST v1.1 every 9 wk. Survival was assessed every 2 mo. Primary end points were OS and PFS in the TPS ≥50% and ≥1% populations. Pembro doses were pooled for this prespecified analysis. Results: Enrollment was based on archival samples in 456 pts (44%) and new samples in 578 (56%). Median (range) time between sample collection and PD-L1 assessment was 250 d (3-2510) and 11 d (1-371), respectively. PD-L1 TPS was ≥50% for 40% of archival and 45% of new samples. Archival samples were used in 48% of 222 pts with squamous and 43% of 724 pts with nonsquamous histology. Pembro significantly improved OS for TPS ≥50% and ≥1% in pts enrolled based on new and archival samples (Table). The PFS benefit pembro vs doce was similar in archival and new samples (Table). Conclusions: Pembro provided superior OS over doce regardless of whether new or archival samples were used to assess PD-L1. Incidence of PD-L1 TPS ≥50% was similar in archival and new samples. These data suggest a new biopsy may not be required for this predictive assay, sparing pts from risks associated with sample collection and avoiding resource utilization. Clinical trial information: NCT01905657