The majority of patients with advanced NSCLC do not respond to monotherapy with PD-1 axis inhibition, and more robust predictive biomarkers are needed. In this study, we assessed tumor samples from 67 immunotherapy-treated NSCLC cases represented in a tissue microarray, 53 of whom had pre-treatment samples and received monotherapy. Using GeoMx Digital Spatial Profiling (DSP) system (NanoString), we quantified 39 immune parameters simultaneously in four tissue compartments defined by fluorescence co-localization (tumor [panCK+], leucocytes [CD45+], macrophages [CD68+], and non-immune stroma). In the univariate unadjusted analysis, we found 18 markers associated with outcome in spatial context, 5 of which remained significant after multiplicity adjustment and after controlling for clinical confounding factors. In the multivariate analysis, high expression of CD56 and CD4 in the CD45 compartment were significantly associated with all favorable clinical outcomes, whereas high levels of VISTA and CD127 in the tumor compartment were markers associated with immunotherapy resistance. We also validated the DSP finding that high CD56+ immune cell counts in the stroma were predictive for PFS and OS in the same set of patients using multiplex immunofluorescence, strengthening the relevance of NK/NKT cells