Oxygen Therapy in Critical Illness

Precise Control of Arterial Oxygenation and Permissive Hypoxemia

Daniel Stuart Martin, BSc, MBChB, PhD, FRCA, FFICM; Michael Patrick William Grocott, MBBS, MD, FRCA, FRCP, FFICM


Crit Care Med. 2013;41(2):423-432. 

In This Article

Potentially Harmful Effects of Excessive Oxygen and Hyperoxemia

Eponymously named after Paul Bert and James Lorrain Smith, the detrimental effects of hyperbaric oxygen on the central nervous system[32] and pulmonary tissues[33] respectively, were well described in the 19th century. More recently, it has become clear that high concentrations of normobaric oxygen may also be harmful. As the gas exchange interface of the body, it is logical that pulmonary tissue would be one of the tissues at greatest risk of damage from high-inspired oxygen concentrations, and this has been demonstrated in numerous animal and healthy human volunteer studies.[34] The damage caused to pulmonary tissue by excessive oxygen resembles the changes seen in acute respiratory distress syndrome (ARDS); the magnitude of injury is directly related to the concentration of oxygen and duration of exposure.[35,36] Oxygen toxicity is rarely evident when the FIO2 is less than 0.5.[1] As patients with ARDS frequently require an FIO2 greater than 0.5, they are potentially at risk of exacerbating the underlying lung injury. It has been previously reported that positive pressure ventilation with a high FIO2 (0.61–0.93) resulted in specific pathological findings independent of the detrimental effects of the ventilator.[37] Clinically, oxygen toxicity can result in decreased mucociliary transport,[38,39] atelectasis (resulting in ventilation–perfusion mismatching), inflammation, pulmonary edema, and eventually interstitial fibrosis.[35] These pathologies may result in worsening lung function.

Excess oxygen administration is thought to damage tissue through the production of reactive oxygen species (ROS). These oxygen-containing molecules that form covalent bonds with other molecules through their unpaired electrons are produced by the mitochondria during oxidative phosphorylation and serve a number of important biological functions. In excessive concentrations, ROS-mediated oxidative stress can lead to cellular necrosis or apoptosis. Paradoxically, hypoxia can also result in an increase in ROS production (40), and ROS are thought to be key players in the pathobiology of reperfusion injury.[41] An important feature of pulmonary oxygen toxicity is that it is almost impossible to distinguish from damage caused by other lung injury processes.[42] Consequently, it is unclear whether deterioration of lung function during high concentration oxygen therapy is due to worsening of the primary disease process or to oxygen-free radical-induced damage; the administration of high concentration oxygen may be perpetuating lung injury in some patients.

Supranormal arterial oxygenation is also associated with a number of cardiovascular responses such as reduced stroke volume and cardiac output,[43,44] increased peripheral vascular resistance,[43] coronary artery vasoconstriction, and reduced coronary blood flow,[45,46] which may be undesirable in critically ill patients.

A growing body of clinical evidence points to the potential harm of using high concentrations of inspired oxygen in clinical situations where classical teaching and physiological intuition might suggest a beneficial response. A number of examples are outlined below:

  1. Acute myocardial infarction. Two recently published systematic reviews of the use of supplemental oxygen during the management of acute myocardial infarction came to the same conclusion; there is no evidence that oxygen therapy (when compared to air breathing) is of benefit in this setting (47), and it may in fact be harmful, resulting in greater infarct size and increased mortality.[48] While the small number of included studies limited the interpretation of these reviews and none of the original studies obtained a statistically significant result,[49] they highlight provocative data that merits further urgent investigation. The AVOID (Air Verses Oxygen In myocarDial infarction) study (NCT01272713) is currently recruiting patients in order to answer this crucial clinical question.[50]

  2. Acute ischemic stroke. Clinical trial data evaluating the effects of different inspired oxygen levels are even more sparse in acute ischemic stroke. Oxygen therapy may be of benefit if administered within the first few hours of onset, but evidence also exists that it may result in increased harm (higher 1-yr mortality) with continued administration.[51]

  3. Neonatal resuscitation. During the past decade, the practice of resuscitating neonates with 100% oxygen has been challenged and many are now advocating that air should be used for initial resuscitation.[52] Several studies have demonstrated that the use of 100% oxygen during the resuscitation of human neonates may increase mortality, myocardial injury and renal injury, and even be associated with a higher risk of childhood leukemia and cancer.[53] Furthermore, in a manner comparable to an ischemia-reperfusion injury, the use of 100% oxygen in the new born following an asphyxiating perinatal event[54] is thought to result in cerebral damage. Such is the evidence base that resuscitation guidelines in neonates now advise that the initial gas administered for ventilation should be air, and that oxygen should be titrated into the mixture according to clinical response so as to avoid hypoxemia.[55]

  4. Adult resuscitation following cardiac arrest. In a retrospective cohort study of more than 6,000 patients following resuscitation from cardiac arrest, hyperoxemia (defined as a PaO2 > 300 mm Hg [40 kPa]) was associated with a significantly worse outcome than both normoxemia (60–300 mm Hg [8 to 40 kPa]) and hypoxemia (< 60 mm Hg [8 kPa]).[56] The authors of this article concluded that excessive oxygen has harmful potential during adult resuscitation post cardiac arrest, possibly via ischemic reperfusion damage to central nervous tissue.

  5. Critical illness. Limited data are available describing the relationship between arterial oxygenation, morbidity, and mortality in critically ill patients. The complexity of separating "signal" from "noise" in this heterogeneous patient cohort makes this task challenging. Among acute medical emergency admissions there is evidence that low SaO2 is an independent predictor of mortality;[57] however, this relationship is more complicated in established critical illness with sustained hypoxemia. Likewise, the degree to which a reduction in arterial oxygenation can be tolerated in the critically ill is difficult to determine and remains unclear.[58]

The assumption that a higher PaO2 is correlated with improved long-term survival in critically ill patients has no robust evidence in its support.[15] A retrospective study of arterial oxygenation in Dutch intensive care patients who were mechanically ventilated within 24 hrs of admission demonstrated a biphasic relationship between PaO2 and in-hospital mortality.[59] Mean PaO2 in this cohort of more than 36,000 patients was 99.0 mm Hg (13.2 kPa), yet the nadir for unadjusted hospital mortality was just below 150 mm Hg (20 kPa). A similar study of patients in Australia and New Zealand reported a mean PaO2 of 152.5 (±109.5) mm Hg (20.3 kPa), representing supraphysiological levels of oxygenation, with 49.8% of the 152,680 cohort being categorized as hyperoxemic (PaO2 > 120 mm Hg [16 kPa]).[60] In contrast to the Dutch cohort, an association between progressive hyperoxemia and in-hospital mortality was not found after adjustment for disease severity, although hypoxemia was associated with elevated mortality. These conflicting data are limited by the methods used: both studies evaluated the association between the single "worst" (lowest P/F ratio) blood gas within the first 24 hrs of admission to an intensive care unit with in-hospital mortality, without quantifying oxygenation during the rest of the patients' critical illness. The difficulty in inferring a clear message from these studies may, therefore, reflect discordance between the severity of acute hypoxemia and subsequent changes in oxygenation along with the consequent adaptive responses that may occur in critically ill patients. Using arterial blood gas data beyond the first 24 hrs of admission may help more clearly define any association between oxygenation and outcome.

Oxygenation in ARDS

The assumption that elevating arterial oxygenation improves outcomes in patients with hypoxemia secondary to ARDS underpins many studies in this field.[61] However, data from clinical trials in patients with ARDS challenge this assumption and frequently oxygenation and long-term outcome seem unrelated.[62–64] While some studies have reported a relationship between arterial oxygenation and mortality, a systematic review of 101 clinical studies of ARDS concluded that P/F ratio was not a reliable predictor of outcome.[65] A variety of interventions including high frequency oscillatory ventilation, prone positioning, inhaled nitric oxide, and extracorporeal membrane oxygenation have been shown to improve arterial oxygenation in patients with ARDS without yielding an outcome benefit.[14,66] Furthermore, different strategies of mechanical ventilation have led to 1) improved oxygenation but unchanged outcome,[67–69] 2) improved outcome but unchanged oxygenation,[70] and 3) deterioration in oxygenation but unchanged outcome.[71] Taken together, these data do not support the assumption that improved oxygenation has a causative relationship with improved clinical outcomes in patients with ARDS.

Three important considerations relate to this discussion. First, supplemental oxygen is a supportive intervention serving to correct a consequence of the underlying pathophysiology, rather than to treat a cause or reverse a disease process. Second, cellular hypoxia is not a prominent feature of ARDS. Third, death in these studies is rarely due to intractable hypoxemia or respiratory failure, but commonly from the underlying cause of ARDS (e.g., systemic inflammation due to sepsis),[72] 73). Taken together, these data suggest that the underlying assumption that merits testing through adequately powered well-designed clinical trials.