Background: Seven retrospective studies have shown that a greater proportion of infusions of immune checkpoint inhibitors (ICIs) in the morning or at early daytime improved survival in patients (pts) with stage IV cancers. Pembrolizumab (Pembro) binds to PD-1 receptors and targets T(CD8) and other immune cells, whose functions and trafficking are regulated by the circadian timing system. Since 4 courses are needed to reach Pembro steady-state plasma levels, we questioned whether Pembro timing would predict survival outcomes. Methods: Non-small-cell lung cancer (NSCLC) pts received Pembro alone or with chemotherapy (CT) according to PD-L1 status q3 weeks. Infusion timeslots were allocated by the day-hospital coordinators and recorded. The study population was split into two groups according to the median timing of the initial four Pembro infusions. The ‘morning’ group received 2 to 4 infusions starting before this median timing, whilst the ‘afternoon’ group received 0 or 1 infusion before it. Progression-free survival (PFS), and overall survival (OS) were analyzed according to timing group. Results: 97 consecutive pts with stage IV NSCLC (M/F, 68/29; PS 0-1, 97%; 37-87 y.o.) received Pembro as 1^st line for 86% of the pts or 2^nd-3^rd line for 14%. Pembro was administered as a single agent to 41 pts (42%) or with CT to 56 pts (58%). Tumor PD-L1 expression was not assessed for 5 pts, < 1% for 18 pts, 1-49% for 20 pts, and ≥ 50% for 54 pts. Main metastatic sites were bone (43% of the pts), pleura (26%), brain (29%), liver (20%), and adrenal gland (14%). The median infusion time of the initial 4 courses was at 11:45. Pts were retrospectively allocated to the ‘morning’ group (48 pts) or the ‘afternoon’ group (49 pts). Pembro+CT was given to 77 % of the ‘morning’ group pts and to 39% of the ‘afternoon’ group pts (p < 0.001), as expected from the longer infusional Pembro+CT duration. All other pt characteristics were similar between both groups. Pts receiving single-agent Pembro or Pembro+CT displayed non-significant differences in PFS (HR = 0.58, p = 0.08) or OS (HR = 0.81, p = 0.57). PFS did not differ significantly between timing groups (HR = 0.64, p = 0.145). In contrast, OS was significantly longer in the ‘morning’ as compared to the ‘afternoon’ group (HR = 0.41 [95%CI, 0.20-0.82], P = 0.012), with respective 2-year survival rates of 65% and 38% (p = 0.010). Multivariable analysis confirmed that morning dosing of 2-4 of the initial 4 Pembro courses predicted longer OS (HR = 0.28 [0.13-0.64], p = 0.003) irrespective of combination with CT, PD-L1 status, and other clinical features. Conclusions: Pembro was largely more effective on OS in those NSCLC pts receiving 2 to 4 infusions before 11:45 over the initial 3 treatment months. Randomized trials and translational studies are now needed to recommend morning timing for Pembro and possibly other ICIs in NSCLC in order to optimize survival benefits from immunotherapy.