Oxygenation Goals in Critically Ill Patients

Monitoring Oxygenation

• In the field of critical care medicine, the mantra “less is more” is a common guiding principle among clinicians when treating critically ill patients. One area where this philosophy can be applied is oxygenation, which is commonly measured by SpO₂, the percentage of hemoglobin where all the binding sites are occupied by oxygen.¹
• Pulse oximetry measurements can be categorized into hypoxemia (SpO₂ <88%), normoxemia (SpO₂ 90-98%), and hyperoxemia (SpO₂ >98%).²
  • This is typically monitored with a pulse oximeter, which is attached to a patient’s appendage (typically a finger, toe, ear, or nostril) and measures SpO₂ via light absorption through a tissue bed with pulsatile blood flow.¹
• While this method is commonly used because it is quick and noninvasive, SaO₂ is considered the gold standard. It provides a more precise and accurate measure of oxygenation. SaO₂ is calculated via the direct measurement of oxygen-bound hemoglobin from an arterial blood gas, which is more invasive as it requires an arterial blood sample.
• Pulse oximetry measurements can be affected by numerous patient factors, including poor circulation, skin temperature, skin thickness, and skin pigmentation.1
• Indeed, patients with darker skin pigmentation may experience greater rates of unrecognized hypoxemia (SaO₂ < 88% when SpO₂ ≥ 92%) and hyperoxemia (PaO₂ > 150 mmHg when SpO₂ is 92-96%) due to the impacts of skin pigmentation on light absorption through the tissue bed.³
  • Unrecognized hypoxemia has even been associated with higher rates of subsequent organ dysfunction and mortality.
• For this reason, the U.S. Food and Drug Administration recommends that pulse oximetry devices be tested among a diverse patient population across multiple demographics, including race and ethnicity.

Clinical Outcomes

• A growing body of evidence has shown that targeting normoxemia (SpO₂ 90-98%) among critically ill patients improves, or at least maintains, clinically meaningful outcomes.

Cao et al. meta-analysis (2023)⁴

• One of the foremost meta-analyses on this topic was conducted by Cao et al (2023), which encompassed 12 randomized controlled trials (RCTs) and 7,416 patients.
• Although there was heterogeneity in oxygenation targets across studies, they found no significant differences between critically ill patients treated with lower versus higher oxygenation targets in the following parameters.⁴
  • All-cause mortality
  • Serious adverse events
  • Organ failure
  • Need for renal replacement therapy
  • Ventilator-free days
  • ICU length of stay

Oxygen ICU Trial (2016)⁵

• This was a single-center RCT conducted in Italy, which was included in the meta-analysis by Cao et al.
• This group reported an absolute risk reduction in mortality of 8.6% among critically ill patients randomized to a conservative SpO_₂ target (94-98%) versus a higher SpO₂ target (97-100%).
• The strength of these findings inspired several other groups to design similar randomized trials of oxygenation targets.
  • While none of the oxygen target RCTs that followed the Oxygen-ICU trial replicated the mortality benefit, all demonstrated that targeting normoxemia at least maintained clinical outcomes.

ICU-ROX Trial (2020)⁶

• The ICU-ROX trial (2020) found no significant difference in ventilator-free days or 180-day mortality between conservative (SpO₂ 90-97%) and liberal (no specific SpO₂ goal) oxygenation strategies in 965 mechanically ventilated patients.

HOT-ICU Trial (2021)⁷

• This multi-center RCT conducted in Europe also reported no difference in mortality among 2,928 patients with acute hypoxemic respiratory failure (AHRF) who were randomized to a target PaO₂ of 60 mmHg versus 90 mmHg.

PILOT Trial (2022)⁸

• Similarly, this group found no difference in 28-day mortality or ventilator-free days among 2,541 mechanically ventilated patients assigned to low (88–92%), intermediate (92–96%), or high (96–100%) SpO₂ targets.

ICONIC Trial (2023)⁹

• The ICONIC trial also reported no significant difference in 28-day mortality among 664 mechanically ventilated intensive care unit (ICU) patients randomized to SpO₂ targets of 91–94% versus 96–100%.

UK-ROX Trial (2025)¹⁰

• Finally, the UK-ROX trial (2025), the largest oxygen-target trial to date, randomized 16,500 mechanically ventilated patients to receive either the lowest possible fraction of inspired oxygen to maintain SpO₂ ≥ 90% or usual care.
• Targeting normoxemia and minimizing oxygen exposure did not decrease 90-day all-cause mortality, the primary outcome.
• There were no differences in any other clinically meaningful outcomes.

Trial Oxygenation Goal Differences

• The previously mentioned 2023 meta-analysis by Cao et al, which included results from the Oxygen-ICU, ICU-ROX, HOT-ICU, and PILOT trials, found no significant differences in morbidity or mortality between patients managed with lower versus higher SpO₂ targets.⁴
• However, the authors noted considerable heterogeneity in the oxygenation goals used across studies.
  • Many of the oxygenation target groups in recent trials were in the normoxemia range.
  • That is, these trials were comparing lower normoxemia targets with slightly higher normoxemia targets.
• These choices likely reflect the growing recognition among clinicians that targeting normoxemia – and avoiding excessive hyperoxemia – is increasingly incorporated into standard clinical practice.
• Dr. Matthew Semler summed these findings up nicely in his discussion of the PILOT Trial: “The results of our trial and three other recent trials suggest that, within the range of SpO₂ values from 90 to 98%, the choice of oxygenation target does not affect clinical outcomes for a broad population of critically ill adults.”⁸

Patient Subgroup Considerations

• Specific subgroups of critically ill patients may experience greater benefits from targeted normoxemia.

Trauma

• Targeting normoxemia among critically ill trauma patients maintains clinical outcomes and reduces the need for supplemental oxygen in critically ill patients without increasing mortality risk.
• The multicenter SAVE-O₂ trial found that critically ill trauma patients receiving targeted normoxemia (SpO₂ 90-96%) weaned to room air more quickly and received less supplemental oxygen.
• Patients receiving targeted normoxemia who were not mechanically ventilated experienced more supplemental oxygen-free days (SOFDs) compared to usual care.
• These results were also notable because there was no significant difference in rates of hypoxemia or mortality between the two groups.²
• Therefore, targeting normoxemia may safely decrease oxygen utilization in settings where oxygen resources are limited.

Stroke and Traumatic Brain Injury (TBI)

• Among patients with stroke or TBI, higher SpO₂ goals do not improve long-term outcomes and may increase the risk of secondary cerebral ischemia or injury.¹¹
• However, hypoxemia can also lead to secondary brain injury and ischemia.
• To balance these competing interests, the American College of Surgeons recommends a SpO₂ target of 94-98% and a PaO₂ target of 80-100 mmHg among TBI patients.

Myocardial Infarction (MI) and Cardiac Arrest

• In patients with MI or cardiac arrest, supplemental oxygen without hypoxia on presentation has been associated with larger infarct sizes and a higher risk of recurrent ischemia, without any mortality benefit.¹¹
• These findings have influenced recent cardiology guidelines, which now advise against supplemental oxygen in post-MI patients with SaO₂ of at least 90%.

Acute Hypoxemic Respiratory Failure (AHRF) and Acute Respiratory Distress Syndrome (ARDS)

• In patients with AHRF, targeting normoxemia has been associated with similar clinical outcomes compared to higher oxygenation targets.
• In the HOT-ICU trial, ICU patients with AHRF were randomly assigned to either a higher oxygenation goal (which was still normoxemic at a PaO₂ of 90 mmHg) or a lower goal (PaO₂ of 60 mmHg).
  • These two groups had no significant difference in 90-day mortality or new serious adverse events (e.g., shock, cerebral or myocardial ischemia).⁷
• Critically ill patients with ARDS should be maintained within conservative oxygenation goals (PaO₂ 55-80 mmHg or SpO₂ 88-94%) while avoiding aggressive oxygenation.^8,9

Individualized Approach to Oxygenation

• While the choice of oxygenation target, within the range of SpO₂ values from 90 to 98%, does not affect clinical outcomes for a broad population of critically ill adults, an individualized approach based on patient characteristics is becoming increasingly favored.¹²
• Buell et al demonstrated this with a machine learning model trained on data from the PILOT and ICU-ROX trials.
  • Their group identified patient subgroups that might benefit more from either conservative or liberal oxygenation targets.
  • The model showed that patient-specific oxygenation goals had a significant effect on 28-day mortality outcomes.¹²
  • Patients who were most likely to benefit from conservative oxygenation goals include patients with¹²:
    • ARDS
    • AHRF (including that caused by COVID-19)
    • Acute brain injuries
    • Cardiovascular disease
  • These benefits are attributed to both the pathophysiologic mechanisms of hyperoxemia-induced tissue damage and observed improvements in clinical outcomes within these groups.
  • On the other hand, patients with sepsis identified by the machine learning model as potentially benefiting from more liberal oxygen targets.¹²
    • While this finding was not statistically significant in the PILOT trial, it remains an area of active investigation and may inform future individualized strategies.⁸

Conclusion

• Targeting normoxemia (SpO₂ 90-98%) at least maintains and may improve clinical outcomes for a broad population of critically ill adults.
• These findings are reinforced by a growing understanding of the physiological harm associated with hyperoxemia (SpO₂ > 98%), including oxidative stress, vasoconstriction, and increased risk of pulmonary and cardiac complications.
• While a single oxygenation target may not suit all patients, the routine application of liberal oxygen therapy in the ICU lacks evidence of benefit and may cause harm in many populations.
• Instead, oxygen therapy should be tailored to each patient’s specific clinical context, with targeted normoxemia used as the default in the absence of compelling indications for higher saturation.
• Patient-centered approaches, including emerging tools such as machine learning models, will be critical in refining these goals based on individual risk profiles.
• Ultimately, the reflexive pursuit of maximal oxygenation should be reconsidered in favor of a more nuanced, evidence-based strategy, one that prioritizes optimal outcomes over one-size-fits-all saturation thresholds.