Reference:  Peters MJ, et al. Conservative versus liberal oxygenation targets in critically ill children (Oxy-picu): a UK multicentre, open, parallel-group, randomised clinical trial. Lancet. December 2023

Guest SkepticDr. Anireddy Reddy is a pediatric intensive care attending physician in the Department of Anesthesiology and Critical Care Medicine at Children’s Hospital of Philadelphia.

Dr. Anireddy Reddy

Case: A 3-year-old girl presents to the emergency department (ED) with fever and respiratory distress. Her parents tell you that she has been sick for almost a week and her symptoms seem to be getting worse. In the past few days, her appetite has decreased, and she is breathing harder. On your exam, you note that she is very tired and can barely keep her eyes open. Her oxygen saturation is 78% on room air. She is breathing at a rate of 70 breaths per minute with diffuse retractions and nasal flaring. Your ED team quickly intubates her and places her on a ventilator. There is some improvement in her oxygen saturation to 92% and she is drawing adequate tidal volumes. The respiratory therapist asks you whether you want to increase the FiO2 to improve her oxygen saturation while awaiting transport to the pediatric intensive care unit.

Background: Oxygen is one of the most prescribed therapies. Sometimes it almost feels like a knee-jerk reaction. We see that pulse ox saturation is low and the first thing we do is apply some form of oxygen. 

But like many interventions, there are potential harms and benefits. Yes, we purposely use the language of potential harms and benefits rather than risk and benefits. Using the term “risk” implies that a negative outcome may or may not happen. While the term “benefits” implies a positive outcome. It’s unbalanced.

Too much oxygen has not been shown to improve outcomes in some conditions and has been associated with harm in others. Our guidelines for the treatment of pediatric acute respiratory distress syndrome also recommend a target saturation of 88-92%.

Clinical Question: What is the optimal target for systematic oxygen in critically ill children receiving invasive ventilation?

Reference:  Peters MJ, et al. Conservative versus liberal oxygenation targets in critically ill children (Oxy-picu): a UK multicentre, open, parallel-group, randomised clinical trial. Lancet. December 2023

  • Population: Children >38 weeks corrected gestational age and younger than 16 years enrolled within 6 hours of being admitted to the Pediatric Intensive Care Unit (PICU) and receiving invasive mechanical ventilation with supplemental oxygen
    • Excluded: known or suspected congenital cardiac disease or sickle cell disease, known pulmonary hypertension, when brain pathology/injury was the primary reason for admission, not expected to survive ICU admission, receiving long-term invasive mechanical ventilation prior, or have end-of-life care plans with limitations in resuscitation
  • Intervention: Conservative oxygenation, defined as a target peripheral oxygen saturation (SpO2) of 88% to 92%.
  • Comparison: Liberal oxygenation, defined as target SpO2 >94%
  • Outcome
    • Primary Outcome: Duration of organ support at 30 days. This was a rank-based endpoint scored 1 to 30, and 31 if the patient died. Major components of organ support included respiratory (invasive and non-invasive), cardiovascular (vasoactives and fluid boluses), renal support, analgesia/sedation, transfusion, neurological, and metabolic support. Details can be found in the supplemental material.
    • Secondary Outcomes: Mortality at PICU discharge, time to liberation from invasive mechanical ventilation, duration of organ support, length of PICU and acute hospital stay, functional status at PICU discharge, incremental costs at 30 days
  • Trial: Pragmatic, multicentre, open-label, randomized controlled trial

Authors’ Conclusions: “Among invasively ventilated children who were admitted as an emergency to a PICU receiving supplemental oxygen, a conservative oxygenation target resulted in a small, but significant, greater probability of a better outcome in terms of duration of organ support at 30 days or death when compared with a liberal oxygenation target. Widespread adoption of a conservative oxygenation saturation target (SpO2 88–92%) could help improve outcomes and reduce costs for the sickest children admitted to PICUs.

Quality Checklist for Randomized Clinical Trials:

  1. The study population included or focused on those in the emergency department. No
  2. The patients were adequately randomized. Yes
  3. The randomization process was concealed. Yes
  4. The patients were analyzed in the groups to which they were randomized. Yes
  5. The study patients were recruited consecutively (i.e. no selection bias). Unsure
  6. The patients in both groups were similar with respect to prognostic factors. Yes
  7. All participants (patients, clinicians, outcome assessors) were unaware of group allocation. No
  8. All groups were treated equally except for the intervention. Unsure
  9. Follow-up was complete (i.e. at least 80% for both groups). Yes
  10. All patient-important outcomes were considered. Yes
  11. The treatment effect was large enough and precise enough to be clinically significant. Yes
  12. Financial conflicts of interest. The authors declared funding from the UK National Institute for Health and Social Care Research for the study, and some authors reported consulting fees and honoraria from other organizations. No significant conflicts of interest were reported​.


They randomized 2,040 children from across 15 UK PICUs: 939 were included in the conservative oxygen group and 933 were included in the liberal oxygenation group. The cohort consisted of 56% females with a median age of 2.5 years.

The median SpO2 was 94% (IQR 93-96) in the conservative group. The median SpO2 was 97% (IQR 96-98%) for the liberal oxygenation group.

Key Results: Among invasively ventilated children receiving supplemental oxygen, conservative oxygenation targets resulted in a shorter duration of organ support at 30 days or death.

Primary Outcome: Duration of organ support or death at 30 days was significantly lower in the conservative oxygenation group compared to the liberal oxygenation group. Probabilistic index of 0.53 (95% CI; 0.5 to 0.55), p=0.04. Adjusted odds ratio 0.84 (95% CI 0.72-0.99).

Secondary OutcomesMortality by day 30 was roughly the same.

The conservative oxygenation group spent a slightly shorter period on invasive mechanical ventilation from the time of randomization (69.6 vs 73.2 hours) P-value=0.03.

Differences in the length of PICU stay and length of hospital stay were not statistically significant. Although the length of PICU stay was a bit shorter in the conservative oxygenation group (4 vs 5 days) with IQR 3-8.

Functional status at PICU discharge measured through the Paediatric Cerebral Performance Category and Paediatric Overall Performance Category were similar.

Healthcare costs up to 30 days was less in the conservative oxygenation group.

Selection Bias:

While many of the exclusion criteria were reasonable, some criteria for this study were subjective. These included the criteria of the patient not expected to survive ICU admission and suspected congenital heart disease or sickle cell disease. Additionally, 40% of the 777 patients who were eligible but did not undergo randomization were due to “clinical decision” or “other reason.” This was not explained in the manuscript. 


While the authors gathered a lot of different types of data, they note that “all other care was determined by the clinical team primarily responsible for the participant’s care.” This is appropriate; however, it does leave the possibility that some of the observed differences may be due to these differences in care rather than attributable to the oxygenation alone.

Protocol Deviation:

There was a higher proportion of patients in the conservative oxygenation group who had protocol deviations (41.4% vs 16.6%). Most of these deviations in the conservative oxygen group involved going above the designated oxygen target range.

We’re not sure why or under what circumstances this happened. This may speak to our discomfort with tolerating lower oxygen saturations. These deviations could also have been due to very legitimate reasons and possibly protective of the patients.

It seems like most of the patients in the conservative treatment group spent the majority of their time with oxygen saturations somewhere in the low to mid 90s. (eFigure 1). They reported a mean SpO2 of 94% (IQR 93-96) in the conservative group compared to 97% (IQR 96-98) in the liberal group.

Figure 2 and the supplemental Etable 6 show that we are comparing SpO2 93-96 with SpO2 97-98.  We’re just showing that the lower end of normal may be beneficial. This is reminiscent of the transfusion threshold trials where the conservative hemoglobin groups were closer to the liberal because no one felt comfortable with a low hemoglobin.


Their primary outcome was defined as the duration of organ support at 30 days or death The definition of organ support was defined by the UK Paediatric Intensive Care Audit and Research Network. This covered multiple organ systems and included things like non-invasive respiratory support, inotropes, bolus IV fluids, dialysis, sedatives, transfusion, and thrombolysis.

But are all these mechanisms of organ support the same? We are argue that there’s a difference between getting a fluid bolus and receiving dialysis. These details of what kind of organ support was slightly prolonged in the liberal oxygen group are not provided in the manuscript. 

The outcomes that we care a lot about such as mortality and functional status at PICU discharge did not differ statistically between the two groups.

Probabilistic Index (PI):

In this study, they reported something called the probabilistic index (PI) which was 0.53 (95% CI 0.5-0.55). This was statistically significant with a p-value of 0.04. This was consistent with their proposed plan for analysis in the clinical trial registry

The PI ranges from 0 to 1. In this study, we can interpret the PI of 0.53 as there is a 53% chance of a better outcome in the conservative oxygenation group. For those interested in learning more about the PI, check out the paper by Acion et al.

The adjusted odds ratio was 0.84 with a 95% CI of 0.72-0.99, narrowly missing that cutoff of 1. When we look at the actual number of days of organ support, there is also quite an overlap. It was 5 (IQR 3 to 9 in the conservative group versus 5 (IQR 3 to 10) in the liberal group. These numbers make the two groups seem much more similar.

Their power calculations need to be revised.  They assumed a mortality of 7% but then excluded a lot of the patients who die in the PICU (TBI, post-arrest, stroke, congenital heart disease, pulmonary HTN) such that the observed mortality rate is closer to 3%.  This has a significant impact on the sample size needed to observe an effect.

Comment on Authors’ Conclusion Compared to SGEM Conclusion: We agree with the overall message that we can likely tolerate more conservative oxygen saturations. However, we do not think there is strong enough evidence to recommend an optimal target saturation of SpO2 88-92%.

SGEM Bottom Line: The optimal oxygen saturation for invasively ventilated children is not known. Lower oxygen saturation targets may be tolerated.

Case Resolution: You check to make sure the endotracheal tube is in the correct position and the ventilator equipment. After finding no obvious errors, you tell the respiratory therapist that you are okay with the current ventilator settings and oxygen saturation in the low to mid-90s.

Clinical Application:

We have alarms for low oxygen saturations. Should we implement one for high oxygen saturations?

In the world of alarm fatigue, the move seems to be to reduce alarms and, indeed, get rid of monitoring sooner.  In Dr. Reddy’s PICU, a desaturation does not alarm a nurse’s phone unless it is 88% for at least one minute.  On the floors of her hospital, there is an effort to remove continuous pulse ox monitoring altogether.  

The subset of patients for whom we are most adherent to a conservative oxygen goal is Acute Respiratory Distress Syndrome (ARDS).  The “harm” of achieving a higher goal feels more tangible – it’s not invisible free radicals causing oxidative stress, but rather that PIP of 45 that is causing barotrauma and the risk of pneumothorax.  Allowing permissive hypercarbia and hypoxia allows clinicians to decrease ventilator settings (and potentially shorten the duration of mechanical ventilation).

There are also concerns that oxygen saturation in darker-skinned individuals may be less reliable.

It is an issue of both generalizability of the study and raises how a well-meaning trial could potentially exacerbate health disparities.  We give credit to the investigators that they were able to run a trial during COVID-19 that is representative of the children seen in UK PICUs. However, they even acknowledge in the supplement that critical illness disproportionately affects racial minorities.  If this is the case, then outcomes stratified by race should be reported to examine if disparities exist. 

There are multiple adult and pediatric studies which show the SpO2 over-estimates the true PaO2 (arterial oxygen) in patients with darker skin.  Practically, this means that the child who’s on the conservative oxygen arm with a SpO2 of 88% may have an arterial SpO2 of 86%, which could lead to inadequate end-organ oxygen delivery.  It may be that the patients who experience the adverse effects of lactic acidosis, seizures, worsening functional status, etc. are those with darker skin.  This is a fundamental problem in the way we care for critically ill children and, until the technology for pulse ox improves, we must be mindful of this difference. 

What do I tell the respiratory therapist?

Please keep the SpO2 in the 90s until you get to the PICU where we can monitor them closely.

Remember to be skeptical of anything you learn, even if you heard it on the Skeptics Guide to Emergency Medicine.