Abstract Purpose: Immune checkpoint inhibitors (ICI) have revolutionized the treatment of non–small cell lung cancer (NSCLC), but predictive biomarkers of their efficacy are imperfect. The primary objective is to evaluate circulating immune predictors of pembrolizumab efficacy in patients with advanced NSCLC. Experimental Design: We used high-dimensional mass cytometry (CyTOF) in baseline blood samples of patients with advanced NSCLC treated with pembrolizumab. CyTOF data were analyzed by machine-learning algorithms (Citrus, tSNE) and confirmed by manual gating followed by principal component analysis (between-group analysis). Results: We analyzed 27 patients from the seminal KEYNOTE-001 study (median follow-up of 60.6 months). We demonstrate that blood baseline frequencies of classical monocytes, natural killer (NK) cells, and ICOS+ CD4+ T cells are significantly associated with improved objective response rates, progression-free survival, and overall survival (OS). In addition, we report that a baseline immune peripheral score combining these three populations strongly predicts pembrolizumab efficacy (OS: HR = 0.25; 95% confidence interval = 0.12–0.51; P < 0.0001). Conclusions: As this immune monitoring is easy in routine practice, we anticipate our findings may improve prediction of ICI benefit in patients with advanced NSCLC.