22. Conservative vs. Liberal Oxygen Targets in Critical Illness

Quick Recap

Cross-cutting protocol — addresses supplemental oxygen targeting in critically ill patients, an everyday ICU decision with a genuinely unsettled, multi-trial evidence base spanning single-center enthusiasm, a trial stopped early for harm in the opposite direction, and the largest, most recent (2025) trial finding no benefit at all. This protocol treats the full arc honestly rather than defaulting to either "hyperoxia is dangerous" or "conservative targets are safe" as a settled conclusion.

1. Definition

Liberal oxygen therapy: targeting a higher oxygen saturation/PaO2 range, historically reflecting common ICU practice of maintaining generous oxygen reserve (e.g., SpO2 targets in the mid-to-high 90s, or unrestricted supplemental oxygen titration) — driven by an intuitive, longstanding clinical instinct that more oxygen reserve is protective against unrecognized hypoxic episodes.

Conservative oxygen therapy: deliberately targeting a lower oxygen saturation/PaO2 range (specific thresholds vary by trial, e.g., SpO2 88–92% or 90–97%, or specific PaO2 ranges), based on the biological rationale that hyperoxia itself carries genuine physiological risk — increased oxidative stress and inflammation, adverse hemodynamic effects (including reduced cardiac output and coronary/cerebral vasoconstriction), absorption atelectasis, and augmented lung injury when combined with mechanical ventilation — not merely "unnecessary excess" but a physiologically active state with its own potential for harm.

The central, unresolved tension this entire evidence base reflects: both hypoxia and hyperoxia have well-documented potential for cellular and organ injury — this is not a question of one strategy being obviously safe and the other obviously risky, but a genuine search for the optimal point along a continuum where both extremes carry real, biologically plausible harm.

2. Pathophysiology

Hyperoxia's proposed mechanisms of harm: increased production of reactive oxygen species and oxidative stress, direct vasoconstrictive effects (particularly relevant to coronary and cerebral circulation, where reduced blood flow despite adequate oxygen content could theoretically worsen regional ischemia in vulnerable territories), reduced cardiac output, absorption atelectasis (particularly with high FiO2), and — specifically relevant to mechanically ventilated patients — augmented ventilator-induced lung injury when hyperoxia and mechanical stretch interact, a mechanism with specific experimental support.

Hypoxia's risks, the countervailing consideration: inadequate oxygen delivery risks tissue and organ injury through the more familiar, intuitive pathway — and conservative oxygen strategies inherently carry a real risk of unintended, undetected hypoxic episodes between intermittent monitoring checks, particularly in unstable or rapidly changing patients, a genuine, quantifiable safety concern that has emerged as a recurring theme across this trial literature (Section 11).

Why population and underlying condition may modify the optimal target: patients with acute hypoxic brain injury may have a specific, biologically plausible vulnerability to hyperoxia-related cerebral vasoconstriction worsening secondary injury, while patients with ARDS undergoing mechanical ventilation may have a specific vulnerability to the hyperoxia-ventilation interaction described above — these population-specific mechanistic considerations partly explain why trial results have not converged on a single, uniform answer across all critically ill patients (Section 11).

3. Immediate Stabilization (ABCDE) — Oxygen Targeting as Part of Breathing Management

Not a standalone acute stabilization scenario; oxygen target selection sits within the Breathing component of critical care management, applicable across nearly all ICU patients:

Checklist:

4. Focused History

• Underlying reason for respiratory failure/oxygen requirement
• ARDS status specifically (relevant given LOCO2's population-specific harm signal, Section 11)
• Acute hypoxic brain injury status specifically (relevant given ICU-ROX's population-specific benefit signal, Section 11)
• Sepsis status (relevant given a specific harm-signal concern raised in an ICU-ROX substudy, Section 11)
• Cardiac history, particularly coronary artery disease (relevant to the theoretical coronary vasoconstriction concern with hyperoxia)

5. Comprehensive System-wise Examination

• Respiratory: oxygenation status, work of breathing, current FiO2/flow requirement
• Cardiovascular: relevant to the theoretical hemodynamic effects of hyperoxia (reduced cardiac output, vasoconstriction) discussed in Section 2

POCUS integration: not a primary component of the oxygen-targeting decision itself.

6. Syndrome Identification — Reframed as Population-Specific Target Classification

• General mechanically ventilated ICU patient: current largest, most rigorous trial evidence (UK-ROX) does not support a mortality benefit from conservative targeting; a moderate, physiologically reasonable target (avoiding both clear hypoxia and clear hyperoxia) rather than a strict, aggressively low conservative target is the more defensible current approach
• ARDS specifically: LOCO2's specific, concerning secondary findings (increased 90-day mortality and mesenteric ischemia with conservative targeting, Section 11) argue against an aggressively conservative oxygen strategy in this population — a genuinely important, population-specific caution distinct from the general ICU conclusion
• Acute hypoxic brain injury: ICU-ROX's own population-specific subgroup finding suggested potentially beneficial outcomes with conservative oxygen strategy in this specific group — though this remains a subgroup finding requiring appropriately cautious interpretation, it represents the most consistent population-specific signal favoring conservative targeting across this literature
• Sepsis: an ICU-ROX substudy specifically found a signal for higher mortality with conservative oxygenation in septic patients — a population-specific caution in the opposite direction from the acute brain injury signal, reinforcing that no single, uniform target is supported across all critically ill subpopulations

7. Differential Diagnosis — Not Applicable

8. Severity/Risk Assessment

Frequency of hypoxic episodes: a genuine, quantified safety concern specifically identified in UK-ROX's conservative arm — increased hypoxic episodes were observed despite the trial's overall null mortality finding, representing a real, measurable cost of the conservative strategy that should be weighed explicitly against any theoretical hyperoxia-avoidance benefit, particularly since the theoretical benefit did not materialize as a demonstrated mortality reduction in this largest, most recent trial.

Monitoring precision limitations: LOCO2's own investigators specifically noted that the frequency of arterial blood gas sampling may have caused underdetection of hypoxemia between samples, and that SpO2 values may lack the precision required to avoid unrecognized hypoxic events in high-risk patients — a genuine, acknowledged methodological limitation affecting how confidently any given trial's conservative arm can be assumed to have avoided clinically significant hypoxic exposure, relevant to interpreting the entire literature's safety comparisons.

9. Investigations

Immediate bedside: continuous SpO2 monitoring; intermittent arterial blood gas sampling as needed, with explicit recognition of the monitoring-precision limitations noted above

10. Point-of-Care Ultrasound — Not Applicable

11. Evidence-Based Management — The Full, Genuinely Unresolved Evidence Arc

Early, Single-Center Enthusiasm — Oxygen-ICU (2016)

• Oxygen-ICU trial (Girardis et al., JAMA 2016): a single-center trial comparing conservative (SpO2 94–98% or PaO2 70–100 mmHg) vs. conventional oxygen therapy — found a striking mortality benefit for the conservative strategy
• Widely acknowledged methodological limitations: the trial has been specifically noted to have "several problems with methodology," and a subsequent influential meta-analysis (IOTA) suggesting hyperoxia increases mortality was found to be heavily weighted by this single trial's results — a pattern structurally similar to several other single-center-enthusiasm-then-multicenter-reversal examples throughout this library (cross-reference the Balanced Crystalloids vs. Saline protocol's SMART-to-BaSICS/PLUS arc, and the Glycemic Control protocol's Leuven-to-NICE-SUGAR arc)

ICU-ROX (2020) — The First Large, Multicenter Test — A Genuinely Mixed Result

• ICU-ROX trial (Mackle et al., NEJM 2020, Australia/New Zealand): conservative vs. usual-care oxygen therapy in mechanically ventilated ICU patients — no significant difference in ventilator-free days (the primary outcome)
• Population-specific subgroup signals, in opposite directions: a specific subgroup with acute hypoxic brain injury showed a signal toward better outcomes with conservative oxygenation, while a sepsis-specific substudy found a signal for higher mortality with conservative oxygenation — these two subgroup findings, pointing in opposite directions within the same overall trial, are a genuinely important early indication that no single, uniform oxygen target is likely to be optimal across all critically ill patient populations

LOCO2 (2020) — Stopped Early, With a Concerning Harm Signal in ARDS Specifically

• LOCO2 trial: liberal vs. conservative oxygen therapy specifically in ARDS patients — did not reduce 28-day mortality with conservative oxygen therapy (the primary outcome), but the trial's secondary outcomes showed a concerning signal: increased 90-day mortality and increased rates of mesenteric ischemia with conservative oxygen therapy
• A specific, plausible explanation for why LOCO2 diverged from ICU-ROX's brain-injury-specific benefit signal: because LOCO2 focused specifically on ARDS, it is possible that fewer patients with the specific conditions previously found to potentially benefit from conservative targeting (e.g., acute hypoxic encephalopathy) were enrolled — reinforcing that population selection substantially shapes which direction a given trial's results point
• A specific, genuine practice-pattern confound: the conservative oxygen group in LOCO2 had a lower incidence of prone positioning than the liberal oxygen group — given proning's own separately, robustly established mortality benefit (cross-reference Prone Positioning in Mechanically Ventilated ARDS protocol), this represents a real, potentially outcome-relevant imbalance in co-intervention delivery between arms, complicating clean attribution of LOCO2's harm signal to the oxygen strategy alone rather than partly to this co-intervention disparity
• Practical significance: LOCO2's harm signal, even with these interpretive caveats, was concerning enough to specifically argue against aggressive conservative oxygen targeting in ARDS — a genuinely important, population-specific caution that should not be smoothed over by the more reassuring general-ICU findings from other trials

HOT-ICU (2021) — A Reassuring, Null Result in Acute Hypoxemic Respiratory Failure

• HOT-ICU trial (Schjørring et al.): lower vs. higher oxygenation targets in acute hypoxemic respiratory failure — found no significant difference in 90-day mortality, broadly consistent with and reinforcing ICU-ROX's overall null finding, without replicating LOCO2's specific harm signal

⚠️ UK-ROX (2025) — The Largest, Most Recent, Most Definitive Trial

• UK-ROX trial (Martin et al., JAMA 2025): a large-scale, multicenter, open-label RCT specifically designed to definitively settle the oxygen-target debate for the general mechanically ventilated ICU population — conservative oxygen therapy (SpO2 88–92%) vs. usual/liberal care
• Primary outcome (90-day mortality): no significant difference — conservative oxygen therapy did not reduce 90-day mortality compared to usual care in mechanically ventilated ICU patients
• A genuine, quantified safety concern: increased hypoxic episodes were observed in the conservative arm, a real, measurable cost of the strategy that did not come with a corresponding demonstrated benefit
• Explicit trial limitations acknowledged by independent review: the trial's open-label design, the absence of longer-term functional outcome data (limited to 90-day follow-up, leaving delayed effects of either hypoxia or hyperoxia unassessed), and pre-specified subgroup analyses (sepsis, brain injury) that showed no consistent trends — though with small subgroup sample sizes limiting confidence in these specific subgroup null findings, meaning UK-ROX's overall null result should not be assumed to have definitively resolved the population-specific questions raised by ICU-ROX and LOCO2
• Practical significance: as the largest, most methodologically robust trial specifically designed to answer this question for the general ICU population, UK-ROX represents the current best evidence that conservative oxygen targeting does not confer a mortality benefit and carries a real, measurable increase in hypoxic episode frequency — independent commentary describes this as "challenging the push for conservative oxygen strategies in all ICU patients," while explicitly noting that further studies are needed to refine targets for specific populations given the trial's own subgroup power limitations

Meta-Analytic Synthesis — Genuine, Persistent Heterogeneity

• Multiple recent meta-analyses (including trial sequential analysis methodology specifically designed to assess whether accumulated evidence is sufficient to reject a clinically meaningful effect) have found no clear beneficial or harmful effect of conservative oxygen therapy on mortality when pooling across this trial literature — with trial sequential analysis specifically able to reject a relative risk reduction of 20% or greater for 90-day and longest-follow-up mortality, though the required information size to definitively reject a similar effect for 28-day mortality specifically was not yet achieved even after pooling the available trials — a genuine, acknowledged statistical limitation on how confidently even the pooled evidence can be interpreted for this specific, earlier timepoint
• A separate recent meta-analysis specifically examining liberal vs. conservative strategies found liberal targets conferred a modest ventilator-free-days benefit without significant differences in ICU length of stay or mortality — a genuinely interesting, somewhat counterintuitive process-outcome signal in the opposite direction from the conservative-strategy enthusiasm that originally motivated this entire research program, though this remains one meta-analytic finding among several with somewhat inconsistent conclusions across the broader literature

Practical Synthesis

The oxygen-targeting question, unlike several other topics in this library that have reached a fairly clear, evidence-supported resolution, remains genuinely, substantively unresolved for the general critically ill population — the largest, most recent, most rigorous trial (UK-ROX) found no mortality benefit from conservative targeting and a real increase in hypoxic episodes, arguing against routine, aggressive conservative targeting as a default strategy. However, population-specific signals persist and should not be discarded: LOCO2's harm signal in ARDS specifically argues against aggressive conservative targeting in that population, while ICU-ROX's acute-brain-injury subgroup signal (and the broader mechanistic plausibility of hyperoxia-related cerebral vasoconstriction concern) suggests conservative targeting may still have a role in that specific population, pending further, adequately-powered confirmation. A moderate, physiologically reasonable oxygen target — avoiding both clear hypoxia and clearly excessive hyperoxia, rather than pursuing either an aggressively conservative or an unrestricted liberal strategy — represents the most defensible current default for the general ICU population, with specific population-based deviations (more caution regarding conservative targets in ARDS; consideration of conservative targets in acute brain injury) individualized per the evidence above rather than applying a single uniform target across all critically ill patients.

12. Organ Support

Interacts directly with the ARDS protocol's lung-protective ventilation framework and Prone Positioning in Mechanically Ventilated ARDS protocol (given LOCO2's specific proning-imbalance confound), and with relevant Neurology System protocols for acute brain injury-specific oxygenation considerations.

13. Disease-Specific Therapy — Not Applicable

This is a titration-target protocol rather than a pharmacotherapy one.

14. Consultation Matrix

15. Monitoring Framework

• Clinical: continuous SpO2 monitoring with adequate frequency to detect both hypoxic and hyperoxic excursions, given the genuine, acknowledged monitoring-precision limitations identified across this trial literature
• Escalation triggers: hypoxic episodes occurring under a conservative target strategy → reassess target appropriateness for the individual patient rather than assuming the strategy is uniformly safe

16. ICU Bundle Checklist

17. Complications

Of conservative oxygen strategies:

• Increased hypoxic episodes, a real, quantified finding in UK-ROX specifically
• Increased 90-day mortality and mesenteric ischemia, specifically in the ARDS population per LOCO2
• Possible increased mortality signal in sepsis, per an ICU-ROX substudy (subgroup finding, requiring cautious interpretation)

Of liberal/hyperoxic strategies:

• Theoretical, mechanistically plausible harms: oxidative stress, hemodynamic effects, absorption atelectasis, augmented ventilator-induced lung injury — not consistently confirmed as clinically significant across the largest, most rigorous recent trials (UK-ROX, HOT-ICU) for the general ICU population

Prevention: individualized, population-appropriate target selection per Section 6; adequate monitoring frequency regardless of chosen strategy

18. Escalation & De-escalation — Not Directly Applicable

19. ICU Discharge Criteria — Not Directly Applicable

20. Documentation & Medicolegal Checklist

• Oxygen target selected and rationale documented, particularly for population-specific deviations from a general moderate target (ARDS, acute brain injury)
• Any hypoxic or hyperoxic episodes documented

21. Key Guidelines

• No major guideline has issued a single, universally mandated oxygen target for the general critically ill population, consistent with the genuine, ongoing uncertainty reflected in this evidence base, particularly following UK-ROX's 2025 null result for the general ICU population alongside persistent, unresolved population-specific signals (ARDS, acute brain injury, sepsis)

22. Landmark Trials

23. Controversies

• This remains one of the more genuinely, substantively unresolved questions in this entire library, in contrast to several other topics where a definitive trial or clear pooled signal has emerged — UK-ROX, despite being the largest and most rigorous trial to date, explicitly could not adequately power its own population-specific subgroup analyses (sepsis, brain injury), meaning the population-specific signals from ICU-ROX and LOCO2 have not been definitively confirmed or refuted by the most recent, largest evidence.
• The LOCO2 proning-imbalance confound deserves explicit acknowledgment: the lower proning rate in LOCO2's conservative-oxygen arm represents a genuine, potentially outcome-relevant co-intervention disparity that complicates clean attribution of the trial's harm signal to oxygen strategy alone — this protocol treats this as a meaningful interpretive caveat rather than dismissing or ignoring it, consistent with this library's general practice of engaging honestly with confounds in discordant evidence rather than selectively citing convenient findings.
• Monitoring precision as a genuine, underappreciated limitation across this entire literature: LOCO2's own investigators explicitly noted that intermittent ABG sampling frequency and SpO2 precision limitations may have caused underdetection of hypoxemia between checks — this is a structural limitation potentially affecting how confidently any trial in this literature can claim its "conservative" arm avoided clinically significant hypoxic exposure, a methodological point relevant to interpreting the entire evidence base, not just LOCO2 specifically.
• The opposite-direction meta-analytic signal (liberal targets → modest ventilator-free-days benefit) is a genuinely underappreciated, somewhat counterintuitive finding relative to the hyperoxia-avoidance enthusiasm that originally motivated this entire research program — worth holding as a further indication that the field's working assumption (conservative = safer) has not been as consistently borne out by the accumulated rigorous trial evidence as the original mechanistic rationale and early single-center data suggested.

24. References

1. Girardis M, Busani S, Damiani E, et al. Effect of conservative vs conventional oxygen therapy on mortality among patients in an intensive care unit: the Oxygen-ICU randomized clinical trial. JAMA. 2016;316(15):1583-1589.
2. ICU-ROX Investigators and the Australian and New Zealand Intensive Care Society Clinical Trials Group; Mackle D, Bellomo R, Bailey M, et al. Conservative oxygen therapy during mechanical ventilation in the ICU. N Engl J Med. 2020;382(11):989-998.
3. Barrot L, Asfar P, Mauny F, et al; LOCO2 Investigators and REVA Research Network. Liberal or conservative oxygen therapy for acute respiratory distress syndrome. N Engl J Med. 2020;382(11):999-1008.
4. Schjørring OL, Klitgaard TL, Perner A, et al; HOT-ICU Investigators. Lower or higher oxygenation targets for acute hypoxemic respiratory failure. N Engl J Med. 2021;384(14):1301-1311.
5. Martin DS, Gould DW, Shahid T, et al. Conservative oxygen therapy in mechanically ventilated critically ill adult patients: the UK-ROX randomized clinical trial. JAMA. 2025;334:398-408.
6. Conservative versus liberal oxygen therapy in relation to all-cause mortality among patients in the intensive care unit: a systematic review with meta-analysis and trial sequential analysis. Med Intensiva. 2023.
7. Comparative efficacy of liberal vs. conservative oxygenation strategies in adult ICU trial populations: a random-effects meta-analysis. Acad Med Surg. 2025.
8. Conservative versus liberal oxygen therapy in mechanically ventilated patients: a systematic review with meta-analysis. Can J Respir Ther. 2025.
9. Panwar R, Hardie M, Bellomo R, et al. Conservative versus liberal oxygenation targets for mechanically ventilated patients: a pilot multicenter randomized controlled trial. Am J Respir Crit Care Med. 2016;193(1):43-51.
10. The Washington Manual of Critical Care, 4th ed. 2025 — oxygen therapy and mechanical ventilation chapters.
11. ICU Protocols: A Step-wise Approach, 2nd ed. — relevant respiratory support content.