Research in context
Evidence before this study
Supplemental oxygen is administered to millions of acutely unwell patients around the world every day. Although oxygen can save the lives of patients with severe hypoxaemia, mechanistic and observational studies suggest that excessive oxygen exposure is common in current clinical practice and could have adverse consequences.
We searched MEDLINE, Embase, CENTRAL and the WHO International Clinical Trials Registry, without language restrictions, from inception to Oct 25, 2017, for randomised controlled trials comparing liberal versus conservative oxygen therapy in acutely ill adults. We excluded studies limited to patients with chronic respiratory diseases or psychiatric disease, patients on extracorporeal life support, and patients treated with hyperbaric oxygen therapy. Specifically, previous meta-analyses of observational studies in critically ill patients suggested an association between hyperoxia and increased in-hospital mortality after cardiac arrest, traumatic brain injury, and stroke, but were limited by inconsistency, risk of bias, and the absence of randomised controlled trials. Meta-analyses of randomised controlled trials comparing liberal versus conservative oxygen therapy in the acute myocardial infarction (four trials) and perioperative settings (eight trials) yielded low-quality overall estimates for mortality because of inconsistency and imprecision. We also identified one systematic review of randomised controlled trials assessing normobaric oxygen therapy for stroke, but this study is at the protocol stage. No studies have systematically reviewed all the available randomised controlled trials for these various conditions.
Added value of this study
This systematic review and meta-analysis of more than 16 000 patients across a broad range of acute illnesses is the first study to provide high-quality evidence that excessive supplemental oxygen can be life-threatening. To the best of our knowledge, this is the most comprehensive systematic review on this topic to date. We found high-quality evidence that liberal oxygen therapy increased the relative risk of in-hospital mortality and mortality at 30 days and at longest follow-up, without any significant improvement in other patient-important outcomes, such as disability, risk of hospital-acquired pneumonia, risk of hospital-acquired infections, or length of hospital stay. These findings are distinct from the widespread view that liberal oxygen therapy for acute illnesses is harmless.
Implications of all the available evidence
Our findings have several potential implications for health-care providers, policy makers, and researchers. In view of the paucity of robust evidence and comprehensive knowledge syntheses, practice guidelines and medical directives on oxygen therapy for acute illnesses have been inconsistent. Our results provide much needed clarification, reporting high-quality evidence that a liberal oxygen strategy increases mortality among a broad range of acute illnesses. Moreover, the dose-response relationship between oxygen saturation and mortality risk highlights the need to implement upper limits of acceptable oxygen saturation for safe oxygen supplementation in patients under the care of emergency personnel, nurses, allied health, and clinicians. Future research is required to identify the precise oxygen strategies that maximise benefit and minimise harm. In view of the global burden of disease and the routine use of oxygen worldwide, the findings of this meta-analysis have immediate and important implications.