How to assess the dangers of hyperoxemia: methodological issues
© BioMed Central Ltd 2011
Published: 30 June 2011
We read with interest the paper by Bellomo and colleagues  in the previous issue of Critical Care. We agree with the authors' final conclusion that there should not be a deliberate policy to decrease the fraction of inspired oxygen (FiO2) in the absence of accurate and reliable oximetry. In the developed world, accurate and reliable oximetry and blood gas results are ubiquitous in ICU settings, the context of their study. However, we have issues with the methods used to come to the conclusion that hyperoxaemia has only a weak relationship with mortality.
If one wishes to show no association of hyperoxemia with outcomes, the best approach is to pick the lowest level of arterial partial pressure of oxygen (PaO2). By analogy, if one wished to assess the risk of speeding prior to traffic accidents, one would not look at the lowest speed or the speed at impact; one would, ideally, look at peak speed or average speed. Kilgannon and colleagues  used the first blood gas measurement in the ICU and found that hyperoxemia (PaO2 of at least 300 mm Hg) was associated with increased risk of death in the hospital. de Jonge and colleagues  looked at mean PaO2 of mechanically ventilated patients in the first 24 hours in ICUs and also reported an increased risk of death in patients with hyperoxemia.
We are also concerned that the conclusion of the study relates to hyperoxemia when defined as a PaO2 of greater than 400 mm Hg whereas the study objective was to analyze the risk of death if the PaO2 was at least 300 mm Hg. We understand that the authors did find excess mortality in their intended study group (and in those with a PaO2 of greater than 200 mm Hg) even after adjustment for illness severity. We are concerned that their negative conclusion is based on a different (and smaller) post hoc subset of patients with a PaO2 of greater than 400 mm Hg. By contrast, Kilgannon and colleagues  re-analyzed their data and reported a clear dose response with lowest hospital mortality in the PaO2 range of 60 to 99 mm Hg and they reported a 24% increase in mortality risk for every 100 mm Hg increase in PaO2.
Rinaldo Bellomo, Michael Bailey and Alistair Nichol (on behalf of the Study of Oxygen in Critical Care Group)
We thank O'Driscoll and Howard for their letter. We share the view that pulse oximetry monitoring in cardiac arrest patients is desirable. However, we wish to emphasize that our methods simply sought to replicate those of Kilgannon and colleagues  given the characteristics of our national database. The view that the PaO2 associated with the highest A-a gradient  is a poorer marker of the chance of exposure to hyperoxemia than the first blood gas measurement  was not supported by our findings. In a sample of 100 patients with detailed arterial blood gas information, the worst PaO2 was significantly more representative of mean PaO2 than was the first PaO2 in the ICU (P < 0.0001 for the first 24, 48, or 72 hours), as reported in our paper. In terms of the speeding analogy, the slowest speed achieved by a driver tells us a lot more about his or her mean speed (the speed that may well matter the most) than does the highest speed in the first 10 minutes. Moreover, our conclusions were not based on the lack of effect at a PaO2 of greater than 400 mm Hg (a parameter chosen by Kilgannon and colleagues ) but on a considered assessment of all observations. We advise caution in over-interpreting data from a study in which no adjustment for illness severity was possible and more than 30% of data were missing . Similarly, the study of de Jonge and colleagues  requires validation. We are currently conducting such an investigation. Irrespective of views and interpretations, we believe it is high time to investigate oxygen therapy in a more systematic fashion. We have begun to do so [1, 5, 6] and intend to persevere.
intensive care unit
arterial partial pressure of oxygen.
- Bellomo R, Bailey M, Eastwood GM, Nichol A, Pilcher D, Hart GK, Reade MC, Egi M, Cooper DJ, the Study of Oxygen in Critical Care (SOCC) Group: Arterial hyperoxia and in-hospital mortality after resuscitation from cardiac arrest. Crit Care 2011, 15: R90. 10.1186/cc10090PubMed CentralView ArticlePubMed
- Kilgannon JH, Jones AE, Shapiro NI, Angelos MG, Milcarek B, Hunter K, Parrillo JE, Trzeciak S, Emergency Medicine Shock Research Network (EMShockNet) Investigators: Association between arterial hyperoxia following resuscitation from cardiac arrest and in-hospital mortality. JAMA 2010, 303: 2165-2171. 10.1001/jama.2010.707View ArticlePubMed
- de Jonge E, Peelen L, Keijzers PJ, Joore H, de Lange D, van der Voort PH, Bosman RJ, de Waal RA, Wesselink R, de Keizer NF: Association between administered oxygen, arterial partial oxygen pressure and mortality in mechanically ventilated intensive care unit patients. Crit Care 2008, 12: R156. 10.1186/cc7150PubMed CentralView ArticlePubMed
- Kilgannon JH, Jones AE, Parrillo JE, Dellinger RP, Milcarek B, Hunter K, Shapiro NI, Trzeciak S, on behalf of the Emergency Medicine Shock Research Network (EMShockNet) Investigators: Relationship Between Supranormal Oxygen Tension and Outcome After Resuscitation From Cardiac Arrest. Circulation 2011, 12: 2717-2722.View Article
- Eastwood GM, Reade MC, Peck L, Jones D, Bellomo R: Intensivists' opinion and self-reported practice of oxygen therapy. Anaesth Intensive Care 2011, 39: 122-126.PubMed
- Eastwood GM, Peck L, Young H, Prowle J, Jones D, Bellomo R: Oxygen administration and monitoring for ward adult patients in a teaching hospital. Intern Med J 2010. [Epub ahead of print]