Objective: Corticosteroids have important effects on intermediate outcomes in critically ill patients, but their effect on survival is unknown. The objective of this meta-analysis was to analyze the effect on mortality of corticosteroids in critical and perioperative settings. Design: A meta-analysis of randomized trials. Setting: PubMed, Embase, BioMed Central, Google Scholar, and the Cochrane Central Register of Controlled Trials were searched to February 1, 2018, for randomized trials comparing corticosteroids with placebo or standard care. Participants: Critically ill or surgical adult patients. Interventions: Corticosteroids compared with placebo or standard care. Measurements and Main Results: A total of 44,553 patients from 135 studies were included. Overall, mortality in the corticosteroid group and in the control group were similar (16% v 16%; p = 0.9). Subanalyses identified a beneficial effect of corticosteroids on survival in patients with respiratory system diseases (9% v 13%; p <0.001) and bacterial meningitis (28% v 32%; p= 0.04), and a detrimental effect on survival in patients with traumatic brain injury (22% v 19%; p <0.001). No differences in mortality were found in patients with cardiac diseases (7% v 6%; p = 0.7), in patients undergoing cardiac surgery (2.8% v 3.2% p = 0.14), and when treatment duration or patient age were considered. Conclusions: This meta-analysis documents the safety of corticosteroids in the overall critically ill population with the notable exception of brain injury patients, a setting where the authors confirmed their detrimental effect on survival. A possible beneficial effect of corticosteroids on survival was found among patients with respiratory diseases and in patients with bacterial meningitis. © 2018 Elsevier Inc.