Sie sind auf Seite 1von 2

Opinion

EDITORIAL

Oxygen in the ICU


Too Much of a Good Thing?
Niall D. Ferguson, MD, MSc

Oxygen gas was discovered and described in the 1770s with exacerbations of chronic obstructive pulmonary disease
by Scheele, Priestley, and Lavoisier and shortly thereafter or with moderate-severe ARDS because of differences in stan-
its therapeutic potential for patients with respiratory ill- dard oxygen administration for these patients. The primary out-
ness was appreciated.1 Oxy- come was predefined as all-cause ICU mortality in the subset
gen has become a mainstay of of patients who remained in the ICU for at least 72 hours after
Related article treatment for acutely ill pa- randomization.
tients, with emphasis fre- The study was stopped early after 480 of a planned 660
quently placed on the importance of avoiding hypoxemia.2 patients had been enrolled, partially related to slow recruit-
Modern clinical practice guidelines for critically ill patients3 ment following ICU bed reductions after an earthquake had
generally target the normal levels for the partial pressure of damaged the hospital, but also following positive results
arterial oxygen (PaO2) and the oxyhemoglobin saturation in ar- from an unplanned interim analysis. These investigators found
terial blood (SaO2), which range from 89 through 100 mm Hg striking results favoring the conservative oxygen group. In the
and 95% through 97%, respectively, depending on patient age.4 modified intent-to-treat population, there were 25 deaths
Avoiding hypoxemia is sensible from a physiological stand- (11.6%) among 216 patients in the conservative oxygen therapy
point because the focus is on maintaining adequate oxygen de- group and 44 deaths (20.2%) among 218 patients in the con-
livery to vital tissues. Oxygen saturation is 1 of only 3 variables, ventional oxygen therapy group, an absolute risk reduction of
along with cardiac output and hemoglobin concentration, that 8.6% (P = .01) for the primary outcome of ICU mortality. The
can be manipulated to achieve this goal. In addition, data from authors found similar results for a secondary outcome of hos-
adults with acute respiratory distress syndrome (ARDS) sug- pital mortality.
gest that lower levels of oxygenation in the intensive care unit These thought-provoking results require close scrutiny
(ICU) are associated with worse long-term cognitive and neu- on several points. It is likely that to some extent, this trial has
ropsychological outcomes.5,6 Furthermore, randomized trials overestimated the true treatment effect of conservative oxy-
of permissive hypoxemia in neonates aimed at reducing reti- gen therapy for a number of reasons. First, there were base-
nopathy of prematurity resulted in higher mortality in the lower line imbalances between groups at the time of randomiza-
oxygen group.7-9 Given these concerns about potential dan- tion, including age, severity of illness, and organ failures,
gers, clinicians often are reassured when the oxygen satura- which favored the conservative oxygen group. Second, the
tion is closer to 100%.10 study was stopped early after a positive finding based on an
Despite these sentiments, however, data increasingly unplanned interim analysis, which is known to increase the
suggest that too much oxygen may also have adverse effects. likelihood of effect overestimation.18,19 Third, although the
That high levels of inspired oxygen can cause lung injury me- overall number of patients recruited was moderate at close to
diated through superoxides and free radical formation has 500, the number of outcome events was actually quite small,
been known for some time.11,12 More recently, toxicity related particularly in the conservative oxygen group, with only 25
to hyperoxia has been observed in a number of other clinical deaths. Even though the P value was significant, the study
situations including cardiac arrest, 13 acute myocardial was still underpowered, which increases the likelihood of a
infarction,14 and brain injury.15 In the critical care unit, a before- false-positive result.20,21
after study also suggested improved outcomes with avoid- The investigators designed their primary analysis using
ance of hyperoxia.16 what they describe as a modified intent-to-treat (ITT) ap-
This background provides the context for the single- proach: analyzing patients according to the group to which they
center randomized clinical trial (RCT) reported by Girardis and were randomly assigned but excluding patients who did not
colleagues17 in JAMA. The authors compared controlled nor- remain in the ICU for at least 72 hours. It should be very clear
moxia (targeting a PaO2 of 70-100 mm Hg or an oxygen satu- that this modified ITT analysis is not equivalent to an ITT analy-
ration by pulse oximetry [SpO2] of 94%-98%) vs usual care, sis. This modified ITT analysis excludes patients based on
which consisted of administering supplemental oxygen at a events after randomization and is extremely problematic be-
fraction of inspired oxygen (FiO2) of 0.4 or higher as needed, cause the intervention may influence who is included in the
targeting a SpO2 of 97% to 100% with a PaO2 up to 150 mm Hg. primary analysis, and this typically leads to larger effect sizes.22
The investigators enrolled patients who were considered likely For example, if conservative oxygen therapy had resulted in
to remain in the ICU for at least 72 hours and excluded those the early death of a number of patients, this would have been

jama.com (Reprinted) JAMA Published online October 5, 2016 E1

Copyright 2016 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ by a Cornell University User on 10/05/2016


Opinion Editorial

overlooked in this analysis plan because those patients were Girardis et al (87 vs 102 mm Hg) relative to their correspond-
excluded. This is in contrast to cases of acceptable postran- ing levels in the ANZICS pilot study (70 vs 92 mm Hg). It is
domization exclusion, which generally involve variables that possible that some of the difference in outcomes between
were present but unknown at the time of randomization.23 these trials was because even in the liberal oxygen arm of the
In this particular case clinicians can be reassured because ANZICS trial, only 22% of time was spent with an SpO2 above
the results of the true ITT analysis (presented in the online 98%, so neither group was exposed to much hyperoxia.
supplementary material) were almost identical to their pri- Given these findings and the possible threats to the valid-
mary analysis plan, with an absolute risk reduction for ICU mo- ity of the trial by Girardis et al,17 what is a clinician to do? It is
rality of 8.7% (P = .009). important to recognize that this study is not a trial of permis-
This trial by Girardis and colleagues contrasts with a sive hypoxemia, which has been proposed but is as yet a
recently published RCT from the ANZICS clinical trials group completely unproven therapeutic strategy. This trial involved
that demonstrated the feasibility and safety of conservative targeting relative normoxia, avoiding both significant desatu-
oxygen administration in a pilot trial of 103 adults.24 This rations and exposure to supraphysiological PaO2. Until the
pilot trial did not, however, show any trend toward improved results of further trials addressing this issue are available,
outcomes with lower oxygen targets. Differences in oxygen- there appears to be little downside in the careful titration and
ation between groups was comparable between these trials, monitoring of supplemental oxygen in the ICU to achieve
but the target and actual PaO2 levels were higher in both con- physiologically normal levels of PaO2 while avoiding poten-
servative and conventional treatment groups in the study by tially dangerous hyperoxia.

ARTICLE INFORMATION 6. Hopkins RO, Weaver LK, Pope D, Orme JF, Bigler Critical Illness: A Systematic Review, Meta-Analysis,
Author Affiliations: Interdepartmental Division of ED, Larson-LOHR V. Neuropsychological sequelae and Meta-Regression of Cohort Studies. Crit Care
Critical Care Medicine and Departments of and impaired health status in survivors of severe Med. 2015;43(7):1508-1519.
Medicine and Physiology, University of Toronto, acute respiratory distress syndrome. Am J Respir 16. Helmerhorst HJF, Schultz MJ, van der Voort PHJ,
Toronto, Canada; Institute of Health Policy, Crit Care Med. 1999;160(1):50-56. et al. Effectiveness and Clinical Outcomes of a Two-
Management, & Evaluation, University of Toronto, 7. Stenson BJ, Tarnow-Mordi WO, Darlow BA, et al; Step Implementation of Conservative Oxygenation
Toronto, Canada; Division of Respirology, BOOST II United Kingdom Collaborative Group; Targets in Critically Ill Patients: A Before and After
Department of Medicine, University Health BOOST II Australia Collaborative Group; BOOST II Trial. Crit Care Med. 2016;44(3):554-563.
Network and Mount Sinai Hospital, Toronto, New Zealand Collaborative Group. Oxygen 17. Girardis M, Busani S, Damiani E, et al. Effect of
Canada; Toronto General Research Institute, saturation and outcomes in preterm infants. N Engl conservative vs conventional oxygen therapy on
Toronto, Canada. J Med. 2013;368(22):2094-2104. mortality among patients in an intensive care unit:
Corresponding Author: Niall D. Ferguson, MD, 8. Schmidt B, Whyte RK, Asztalos EV, et al; the Oxygen-ICU randomized clinical trial. JAMA. doi:
MSc, Toronto General Hospital, 585 University Ave, Canadian Oxygen Trial (COT) Group. Effects of 10.1001/jama.2016.11993
11-PMB-120, Toronto, ON M5G 2N2, Canada targeting higher vs lower arterial oxygen 18. Montori VM, Devereaux PJ, Adhikari NKJ, et al.
(n.ferguson@utoronto.ca). saturations on death or disability in extremely Randomized trials stopped early for benefit:
Published Online: October 5, 2016. preterm infants: a randomized clinical trial. JAMA. a systematic review. JAMA. 2005;294(17):2203-
doi:10.1001/jama.2016.13800 2013;309(20):2111-2120. 2209.
Conflict of Interest Disclosures: The author has 9. Carlo WA, Finer NN, Walsh MC, et al; SUPPORT 19. Bassler D, Briel M, Montori VM, et al; STOPIT-2
completed and submitted the ICMJE Form for Study Group of the Eunice Kennedy Shriver NICHD Study Group. Stopping randomized trials early for
Disclosure of Potential Conflicts of Interest and Neonatal Research Network. Target ranges of benefit and estimation of treatment effects:
none were reported. oxygen saturation in extremely preterm infants. systematic review and meta-regression analysis.
N Engl J Med. 2010;362(21):1959-1969. JAMA. 2010;303(12):1180-1187.
REFERENCES 10. Panwar R, Capellier G, Schmutz N, et al. Current 20. Christley RM. Power and error: increased risk
oxygenation practice in ventilated patients-an of false positive results in underpowered studies.
1. Heffner JE. The story of oxygen. Respir Care. observational cohort study. Anaesth Intensive Care.
2013;58(1):18-31. Open Epidemiol J. 2010:1-4.
2013;41(4):505-514.
2. Martin DS, Grocott MPW. Oxygen therapy in 21. Ridgeon EE, Young PJ, Bellomo R, Mucchetti M,
11. Bryan CL, Jenkinson SG. Oxygen toxicity. Clin Lembo R, Landoni G. The fragility index in
critical illness: precise control of arterial Chest Med. 1988;9(1):141-152.
oxygenation and permissive hypoxemia. Crit Care multicenter randomized controlled critical care
Med. 2013;41(2):423-432. 12. Fisher AB. Oxygen therapy. Side effects and trials. Crit Care Med. 2016;44(7):1278-1284.
toxicity. Am Rev Respir Dis. 1980;122(5 Pt 2):61-69. 22. Abraha I, Cherubini A, Cozzolino F, et al.
3. O’Driscoll BR, Howard LS, Davison AG; British
Thoracic Society. BTS guideline for emergency 13. Kilgannon JH, Jones AE, Shapiro NI, et al; Deviation from intention to treat analysis in
oxygen use in adult patients. Thorax. 2008; Emergency Medicine Shock Research Network randomised trials and treatment effect estimates:
63(suppl 6):vi1-vi68. (EMShockNet) Investigators. Association between meta-epidemiological study. BMJ. 2015;350:h2445.
arterial hyperoxia following resuscitation from 23. Fergusson D, Aaron SD, Guyatt G, Hébert P.
4. Crapo RO, Jensen RL, Hegewald M, Tashkin DP. cardiac arrest and in-hospital mortality. JAMA.
Arterial blood gas reference values for sea level and Post-randomisation exclusions: the intention to
2010;303(21):2165-2171. treat principle and excluding patients from analysis.
an altitude of 1,400 meters. Am J Respir Crit Care
Med. 1999;160(5 Pt 1):1525-1531. 14. Stub D, Smith K, Bernard S, et al; AVOID BMJ. 2002;325(7365):652-654.
Investigators. Air Versus Oxygen in 24. Panwar R, Hardie M, Bellomo R, et al; CLOSE
5. Mikkelsen ME, Christie JD, Lanken PN, et al. ST-Segment-Elevation Myocardial Infarction.
The adult respiratory distress syndrome cognitive Study Investigators; ANZICS Clinical Trials Group.
Circulation. 2015;131(24):2143-2150. Conservative versus liberal oxygenation targets for
outcomes study: long-term neuropsychological
function in survivors of acute lung injury. Am J 15. Helmerhorst HJF, Roos-Blom M-J, mechanically ventilated patients: a pilot multicenter
Respir Crit Care Med. 2012;185(12):1307-1315. van Westerloo DJ, de Jonge E. Association Between randomized controlled trial. Am J Respir Crit Care
Arterial Hyperoxia and Outcome in Subsets of Med. 2016;193(1):43-51.

E2 JAMA Published online October 5, 2016 (Reprinted) jama.com

Copyright 2016 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ by a Cornell University User on 10/05/2016

Das könnte Ihnen auch gefallen