Background: Nivo, afully human IgG4 programmed death-1 immune checkpoint inhibitor, is approved in the US for pts with previously treated metastatic NSCLC and in the EU for pretreated locally advanced or metastatic squamous (SQ) NSCLC, based on results of 2 ph III trials. Methods: Pts with SQ (CheckMate 017) or non-squamous (NSQ; CheckMate 057) NSCLC received nivo 3 mg/kg Q2W or doc 75 mg/m² Q3W (1:1 randomization) until progression or discontinuation due to toxicity/other reasons. The primary objective in each study was OS. Multivariate exploratory analyses of baseline serum cytokines were performed separately in pts with SQ and NSQ NSCLC. A 6:4 training:test validation was used. A SQ-cytoscore derived from evaluable pts in CheckMate 017 and CheckMate 063 (single-arm phase II study of nivo in SQ NSCLC) and a NSQ-cytoscore derived from evaluable pts in CheckMate 057 were generated to quantify the effect of each identified cytokine set on OS (using 18-mo data cutoffs). Cytoscores were defined as high or low based on the median cutoffs. Results: In CheckMate 017, median OS (mOS) was 9.2 vs 6.0 mo with nivo vs doc (18-mo OS: 28% vs 13%; HR: 0.62 [0.48, 0.81]; P= 0.0004). In CheckMate 057, mOS was 12.2 vs 9.4 mo with nivo vs doc (18-mo OS: 39% vs 23%; HR: 0.72 [0.60, 0.88]; P= 0.0009). A greater magnitude of benefit was noted for pts with PD-L1–expressing NSQ NSCLC; PD-L1 expression was neither prognostic nor predictive of benefit in pts with SQ NSCLC. Treatment-related AEs were less frequent with nivo vs doc in both trials. Preliminary results show an association of SQ- and NSQ-cytoscores with OS (Table). Conclusions: Nivo resulted in improved OS and a favorable safety profile vs doc across NSCLC histologies; updated 2-y OS and safety will be presented. Select sets of baseline serum cytokines (details to be presented) were found to be associated with OS benefit in pts with adv SQ and NSQ NSCLC; prospective validation of these results is needed. Clinical trial information: NCT01642004, NCT01673867