Date: January 28th, 2019

Reference:  Chu DK et al. Mortality and morbidity in acutely ill adults treated with liberal versus conservative oxygen therapy (IOTA): a systematic review and meta-analysis. The Lancet 2018.

Guest Skeptic: Dr. Robert Edmonds is an emergency physician in the US Air Force in Virginia. This is Bob’s ninth visit to the SGEM.


Case: You’re working a shift in a rural emergency department when a 68-year-old man presents with a two-day course of worsening cough, shortness of breath, and fever. Their workup reveals a multifocal pneumonia with signs of sepsis.  The patient has an oxygen saturation of 98% on room air and your nurses ask if you want the patient to receive supplemental oxygen.

Background: The liberal use of supplemental oxygen therapy in acutely ill adults has a long history in the hospital, but high-quality therapy supporting its practice is unclear.

Recently, the role of oxygen therapy in non-hypoxic patients has been challenged in myocardial infarction patients, as seen in a number of trials including DETO2X-AMI

We covered the DETO2X-AMI trial on SGEM#192. The SGEM Bottom Line was that the routine administration of supplemental oxygen in patients with suspected or confirmed acute myocardial infarction who are not hypoxic does not appear to provide a patient-oriented benefit.

In the 2015 AVOID study there was some suggestion of increased MI size in the group of STEMI patients that received oxygen at 8 L/min compared to a room air control group. This study expands upon that investigation to patients with other conditions as well.

While supplemental oxygen is undoubtedly beneficial for patients acutely desaturating, in respiratory distress, or suffering from carbon monoxide poisoning just to name a few, there is widespread “indication creep” for this therapy.

In neonatal resuscitation oxygen is treated like a drug that should be appropriately dosed, with careful attention to limit its use to the minimum required amount out of a fear of harm from its excess use.

In acutely ill adults, this same concept is not yet as widespread and liberal administration is still common place.

Clinical Question: Is liberal oxygen therapy vs. conservative oxygen therapy for acutely ill adults effective and safe?

ReferenceChu DK et al. Mortality and morbidity in acutely ill adults treated with liberal versus conservative oxygen therapy (IOTA): a systematic review and meta-analysis. The Lancet 2018.

  • Population: Acutely ill adults (>18 years old) with any condition requiring non-elective hospital admission and the potential to be exposed to supplemental oxygen
    • Exclusions: Studies with patients who are younger than 18, pregnant, limited to patients with chronic respiratory disease, on extracorporeal life support, treated with hyperbaric oxygen or undergoing elective surgery
  • Intervention: A higher oxygen target (liberal group). This was measured by FiO2, PaO2, arterial oxygen saturation measured by blood analysis or peripheral oxygen saturation measured by pulse oximeter.
  • Comparison: The lower oxygen target (conservative group)
  • Outcomes: The authors do not report any one outcome as a primary outcomes, but instead listed morbidity and mortality.
    • Mortality: In-hospital, 30 days, and at the longest follow-up
    • Morbidity: Disability measured by the modified Rankin Scale, risk of hospital-acquired pneumonia, risk of any hospital-acquired infection or hospital length of stay.

Authors’ Conclusions: “In acutely ill adults, high-quality evidence shows that liberal oxygen therapy increases mortality without improving other patient-important outcomes. Supplemental oxygen might become unfavourable above an SpO2 range of 94–96%. These results support the conservative administration of oxygen therapy.”

 Quality Checklist for Therapeutic Systematic Reviews:

  1. The clinical question is sensible and answerable. Yes
  2. The search for studies was detailed and exhaustive. Yes
  3. The primary studies were of high methodological quality. Unsure
  4. The assessment of studies were reproducible. Yes
  5. The outcomes were clinically relevant. Yes
  6. There was low statistical heterogeneity for the primary outcomes. Yes and No
  7. The treatment effect was large enough and precise enough to be clinically significant. Yes

Key Results: The search found 25 randomized control trials to include in the meta-analysis with 16,037 patients suffering from sepsis, critical illness, stroke, trauma, myocardial infarction, cardiac arrest or emergency surgery.

A liberal oxygen strategy increased the risk of death compared with a conservative strategy in hospital, at 30-days, and at longest reported follow-up.

Meta-regression analysis demonstrated that increasing SpO2 was associated with high relative risky of in-hospital ant at longest follow-up mortality.

The morbidity findings (disability measured by the modified Rankin Scale, risk of hospital-acquired pneumonia, risk of any hospital-acquired infection or hospital length of stay) were similar between groups.

1Statistical Power of Systematic Review Meta-Analysis:This study collating multiple negative studies into a positive one. The claim of increased mortality from a liberal oxygen strategy is certainly eye catching. The liberal oxygen group had 283 deaths in 7,555 patients. More than three-quarters (78%) of these deaths are from four of the 19 studies (not all 25 included studies had mortality data).  In each of these four studies, the 95% CI for the relative risk crossed or included the number 1.00 which would make the individual studies not statistically significant. Of these four studies demonstrating a higher incidence of death studies, three of them had a 95% CI that went as low as 0.81 and as high as 1.66; the 95% CI was 1.00-1.78 in the other study of these big four by Girardis et al.  This is of concern as the meta-analysis is effectively turning insignificant studies into something else and is pretty heavily weighted by one particular study which constituted only 3% of the overall meta-analysis population but 26% of all deaths.

2) Variety: There was a wide variety of patients included and a wide variety of protocols used in the individual trials. This can be a potential strength and potential weakness. It could demonstrate a robust negative impact of oxygen on something like mortality. It could also fail to find a small important difference that truly exists for a specific condition because the noise drowns out the signal.

3) Different Modes of Ventilation Matter: The Girardis study that has the most weight (32%) in the authors’ evaluation of mortality is from 2016 and was of patients in critical care that were invasively mechanically ventilated-in sharp contrast to the patients from the other three of these big four higher incidence of death studies (Roffe 2017, Ronning 1999, and Hoffmann 2017) that underwent face-  mask or nasal prong therapy.  It’s arguable that the overall medical condition of an intubated hyperoxic patient may be significantly different than a patient receiving noninvasive oxygen delivery, and this may add concern to the previous point of one study weighing the rest of the analysis heavily.

4) Heterogeneity: This is similar to the second point I was making earlier. While the heterogeneity for mortality was low (zero percent) the heterogeneity for morbidity was high. This makes it more difficult to interpret these results and have confidence in any conclusions.

5) Heterogenous Follow-Up: The two main studies that contributed to the total population of this meta-analysis were the Hoffman and Roffe studies, which contributed 79% of the meta-analysis population.  Hoffman reported 1-year survival and Roffe reported 90-day survival, and its arguable that this significantly impacts the reporting quality of the “mortality at longest follow-up” endpoint.

Comment on Authors’ Conclusion Compared to SGEM Conclusion: We agree that there does not appear to be a benefit to the liberal oxygen therapy approach and high-quality evidence demonstrated an increase in mortality using a liberal oxygen approach. 

SGEM Bottom Line: The goal of oxygen therapy should not usually be 100% in critically ill patients but rather aim for the mid 90’s%.  

Case Resolution: As the patient is breathing easily and has adequate oxygen saturation, I request nursing only supply additional oxygen if the patient begins the have low oxygen saturations.  The patient is admitted to the medical ward on room air.

Dr. Robert Edmonds

Clinical Application: This is another study reminding us that oxygen is a drug and just like any other drug it can have potential benefit and potential harm depending on how and when it is used. Multiple recent studies support a conservative supplemental oxygen therapy approach should be used when treating critically ill patients.

What Do I Tell My Patient? We will be continuously monitoring your body and your vital signs during your hospitalization with us.  It doesn’t appear that additional oxygen will be helpful to you at this time, but if it becomes needed, we’ll supply it at once.

Keener Kontest: Last weeks’ winner was Dr. Aisha Mirza and emergency physican from Edmonton. She knew. Ketamine is the safest agent to use for procedural sedation in children based on the Bhatt et al paper in JAMA Ped 2017. 

Listen to the SGEM podcast on iTunes to hear this the new keener question. If you know the answer send an email to with “keener” in the subject line. The first correct answer will receive a cool skeptical prize.

Other FOAMed Resources:

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