IntroductionWe report the final overall survival (OS) analyses of atezolizumab-carboplatin-paclitaxel (ACP [experimental arm]) and OS data with approximately 39.8 months of median follow-up with atezolizumab-bevacizumab-carboplatin-paclitaxel (ABCP) versus bevacizumab-carboplatin-paclitaxel (BCP) in chemotherapy-naive patients with metastatic nonsquamous NSCLC in the phase 3 IMpower150 study (NCT02366143). MethodsIn this randomized, open-label study (N = 1202), coprimary end points included investigator-assessed progression-free survival and OS in intention-to-treat (ITT) wild-type (WT; no EGFR or ALK alterations) patients. Secondary and exploratory end points included OS in ITT and programmed death-ligand 1 (PD-L1) subgroups defined by the VENTANA SP142 and SP263 immunohistochemistry assays. ResultsAt the final analysis with ACP versus BCP (data cutoff: September 13, 2019; minimum follow-up: 32.4 mo), ACP had numerical, but not statistically significant, improvements in OS (ITT-WT: median OS = 19.0 versus 14.7 mo; hazard ratio = 0.84; 95% confidence interval: 0.71–1.00). OS benefit was sustained with ABCP versus BCP (ITT-WT: 19.5 versus 14.7 mo; hazard ratio = 0.80; 95% confidence interval: 0.67–0.95). Exploratory analyses in the SP142-defined PD-L1 subgroups revealed longer median OS with ABCP and ACP versus BCP in PD-L1–high and PD-L1–positive subgroups; in the PD-L1–negative subgroups, median OS was similar with ACP and ABCP versus BCP. Safety was consistent with that in earlier analyses (data cutoff: January 22, 2018). ConclusionsAt the final IMpower150 OS analysis, ACP had numerical, but not statistically significant, OS improvement versus BCP. Updated data with an additional 20 months of follow-up revealed continued OS improvement with ABCP versus BCP in all patients.