Perez‐Gracia, José Luis

9090 Background: We identified factors associated with better OS for previously treated, PD-L1–expressing advanced NSCLC using data from KEYNOTE-010 (NCT01905657; Herbst et al. Lancet. 2016;387:1540-50), in which pembrolizumab had superior OS over docetaxel. Methods: 1033 patients were randomly assigned 1:1:1 to pembrolizumab 2 or 10 mg/kg every 3 weeks (Q3W) or docetaxel 75 mg/m2 Q3W. Response was assessed per RECIST v1.1 by independent central review. Multivariate analyses were performed using a Cox proportional hazards regression model on OS in the pembrolizumab arms. A set of variable selection methods was applied to 19 baseline demographic and disease characteristics, including smoking status, and identified 7 factors that contributed to OS. Data cut was September 30, 2016. Results: Adjusted hazard ratios (HRs) for the factors in the pembrolizumab arm from the model are shown in the Table. Updated OS with an additional 6 months of follow-up from this data lock for KEYNOTE-010 will be presented. Conclusions: While the overall result of KEYNOTE-010 revealed improved OS with pembrolizumab compared with docetaxel in previously treated patients with PD-L1–positive advanced NSCLC, exploratory, post hoc multivariate analyses showed that some laboratory and tumor characteristics such as nonsquamous histology, normal baseline lactate dehydrogenase (LDH), PD-L1 TPS ≥50%, and wild-type EGFR mutation status were associated with better OS among patients treated with pembrolizumab. Clinical trial information: NCT01905657. [Table: see text]