UK-ROX Proves the U.K. Rocks, But

Let's see what it proved about oxygen in the ICU

More proof: Oasis in 2025 (wikipedia)

High oxygen doses are toxic to the mammalian respiratory system, as noted at least as far back as 1927 in animal testing. In the 1980s, healthy human test subjects breathing 100% oxygen for 24 hours at sea level were observed to develop substernal discomfort, airway erythema and leakage of protein into the alveoli on bronchoscopy. Oxidative damage by free radicals is the dominant putative mechanism of oxygen toxicity, but multiple avenues of injury have been postulated.

Critically ill, mechanically ventilated patients are frequently exposed to days or weeks of high oxygen concentrations. Because these patients already have severe respiratory illness (e.g., pneumonia), any independent contribution of oxygen toxicity to their ongoing respiratory failure or its outcome has thus far been impossible to elucidate.

One key unknown is whether oxygen toxicity exhibits a threshold effect. While tissue injury can clearly occur from prolonged exposures to O2 concentrations over 70%, any meaningful risk from lower concentrations remains hypothetical.

Trial after trial after large randomized trial testing oxygen targets in critically ill patients have been negative, with an inconsistent signal suggesting “conservative” (lower-dose) oxygen therapy might be safer.

Is higher or lower oxygenation better in critical illness? Let's settle this (Review)

But uncertainty has remained, especially with regard to specific subgroups of patients’ potentially variable responses to oxygen therapy. Because hundreds of thousands of people are treated with supplemental oxygen during mechanical ventilation each year, small harms from oxygen toxicity might result in thousands of preventable deaths.

Two super-duper-huge randomized trials were therefore conceived and launched. Results from the smaller of the two mega-trials were recently reported.

UK-ROX

UK-ROX enrolled 16,500 critically ill, mechanically ventilated adults at 97 ICUs in the U.K. between 2021 and 2024. Patients were severely ill: about a third with sepsis, about 10% with hypoxic-ischemic encephalopathy, and most with moderate to severe hypoxemia. They were individually randomized to either a conservative oxygen strategy (targeting SpO2 90% i.e. PaO2 ~60 mm Hg) or usual care (according to clinician preference).

The Brits managed to complete this massive trial over three years at a reported cost of $2.3 million, a miniscule fraction of what such a trial would have cost in the U.S.

This was in large part achieved by bypassing the process of obtaining individual informed consent before enrollment in each patient. Patients were enrolled by default, with consent obtained later from the patient or a surrogate.

So, from the standpoint of facilitating pragmatic research in critical care, the U.K. rocks. But what did we actually learn from UK-ROX?

UK-ROX Results

After 90 days of follow-up, 35.4% of patients randomized to conservative oxygen therapy had died, compared with 34.9% of patients receiving usual care: a 0.5% absolute difference, which was statistically non-significant.

There were no significant differences in secondary outcomes (length of stay, days alive and free of various forms of organ support, etc).

There were no significant differences in adverse events overall.

There were no differences in mortality by oxygen assignment in the predefined patient subgroups: sepsis, hypoxic-ischemic encephalopathy, and brain injury (non-HIE).

UK-ROX was 10 to 20 times the size of the other “large” trials testing oxygen therapy in the ICU.

It was negative.

One would think that should settle this question.

Did it?