14, 13, 11, 30
Liberal oxygen strategy: SaO2 97-100%, oxygenator FbO2 1.0, ventilator FiO2 minimum 0.5
Conservative oxygen strategy: SaO2 92-96%, oxygenator FbO2 0.5-1.0, ventilator FiO2 minimum 0.21
Open-label. Modest size for a mortality endpoint. The 29.5% vs 1.3% deviation rate means achieved separation between arms was imperfect, so a true difference would have been harder to detect — the null is a little weaker than it appears. Primary endpoint (ICU-free days) had a median of 0 in both arms, which limits its discriminating power. Full author list beyond the first three and the precise definition of 'major protocol deviation' not independently checked.
Adults over 18 commenced on VA ECMO in ICU. Registry-embedded, open-label, multicentre, parallel-group RCT; enrolled at 12 hospitals, drawn from 934 VA ECMO patients reported to the registry across 26 hospitals. 149 conservative, 151 liberal. Exclusions: randomisation more than 6 h after ECMO initiation, pregnancy, unwillingness to receive blood products, concurrent oxygen titration study, clinician-specified oxygen targets.
ICU-free days to day 28
Median 0 days (IQR 0-13.7) conservative vs 0 days (IQR 0-13.3) liberal; median treatment effect 0 days (95% CI -3.1 to 3.1). Day 28 mortality 59/149 (39.6%) vs 59/151 (39.1%); day 60 64/149 (43%) vs 62/151 (41.1%). Duration of ECMO and mechanical ventilation, ICU and hospital length of stay, and 6-month functional outcomes all similar. Subgroups consistent (cardiogenic shock, ECPR, age 50 or under vs over 50, SAVE score above/below median, randomisation timing). Critical hypoxaemia, CPR and seizures similar. MAJOR PROTOCOL DEVIATIONS: 44 (29.5%) conservative vs 2 (1.3%) liberal, p less than 0.001. Authors: 'a conservative compared to a liberal oxygen strategy, did not affect the number of ICU-free days to day 28.'
It does NOT license hyperoxia. The liberal arm targeted SaO2 97-100%, not deliberate hyperoxia, so the trial says nothing about running FbO2 1.0 and ignoring the resulting PaO2 — and the observational literature associating hyperoxia with harm after arrest is not overturned by 300 patients. It says nothing about any other VA target: flow, MAP, pulse contour, DO2:VO2, transfusion threshold and venous saturation all remain untested by randomisation.
THE FIRST RANDOMISED TRIAL OF HOW TO SET VA ECMO rather than whether to use it — it corrects Chapter 13's statement that no VA target has randomised support. A conservative oxygen target confers no benefit on any measured outcome. Equally important and more practical: the conservative target is HARD TO DELIVER — nearly a third of that arm could not be held in band, because the oxygenator returns blood at near-maximal saturation and VA mixing physiology offers little fine control. A target a trial team could hold in only 70% of patients should not be enforced rigidly on a ward round.