Is higher or lower oxygenation better in critical illness? Let's settle this (Review)
A set of highly unauthorized meta-analyses
This post is long with multiple forest plots and tables. If you’re reading this in an email, click the title to view it on pulmccm.org or in the Substack app for a better experience. —Ed.
Everyone loves oxygen. But can there be too much of a good thing?
Oxygen has been used as a medical treatment since at least 1885 and is provided to millions of hospitalized people each year, including most ICU patients. But we still don’t know if there’s a “right” dose.
Theoretical harms from excessive oxygen therapy include absorption atelectasis, acute lung injury, inflammation and tissue damage mediated by reactive oxygen species, as well as central nervous system, cerebrovascular and cardiovascular toxicity.
Patients with oxygen-starved tissues (during strokes, myocardial infarction, sepsis) have been hypothesized (but never shown) to benefit from higher oxygen delivery.
Guidelines vary widely on appropriate oxygen targets for various patient populations and none are evidence-based.
Dozens of randomized trials have been performed, and each one suggests something slightly different than the last.
Two massive ongoing randomized trials (or trial platforms), UK-ROX (centered in the U.K.) and MEGA-ROX (based in Australia), are enrolling many patients with many various illnesses (strokes, cardiac arrest, sepsis, etc.) to determine whether certain subgroups of patients benefit from higher or lower oxygen targets.
But these mega-trials won’t provide results for years to come.
Can’t we just settle this now?
1. IOTA
First, consider the largest meta-analysis performed to date: IOTA (Chu et al, Lancet 2018).
IOTA included 25 randomized trials enrolling 16,037 patients with critical illness, sepsis, stroke, trauma, myocardial infarction, cardiac arrest, or emergency surgery. Almost all the trials were published between 2005 and 2017, with only three before 1990.
There was the lowest possible degree of heterogeneity between trials (I-squared was zero).
Almost all trials had mortality as an outcome. Analyzing only the 19 using in-hospital mortality as the outcome, a conservative approach to oxygenation was strongly favored as likely reducing mortality (or a liberal approach was disfavored for increasing mortality): HR 1.21 for mortality with higher oxygenation targets (95% CI 1.03 to 1.43):

The signal from these trials seems crystal clear: liberal or high oxygenation targets appear to be harmful on average.
So why does this question continue to be studied?
Let’s dig deeper.
Keep reading with a 7-day free trial
Subscribe to PulmCCM to keep reading this post and get 7 days of free access to the full post archives.