Oxygen: when is more the enemy of good?

Oxygen is a drug. As a drug, it is unique in that humans are continuously exposed to a low, homeopathic dose from the atmosphere. Clinically, the primary indication for supplemental oxygen is to reverse hypoxemia. This drug is unique in that it is regularly delivered in unknown doses to patients without indications or attempts at titration. The casual use of oxygen by the health community may simply be secondary to the ease of application via the respiratory tract and the misnomer that too much oxygen is clearly better than too little. However, as with many medical therapies, better can be the enemy of good.

The benefit of oxygen is often conflicted with its basic yet forgotten biological harm. Severinghaus and Astrup [1] provide us with a novel and eloquent description of oxygen when they state:

“Oxygen is addicting; in its grip are all the mitochondrial-rich eukaryotes who learned to depend on it during the past 1.4 billion years. This, the first atmospheric pollutant, is the waste product of stromalites (formation of algal plankton), which excreted it at least 2.3 billion years ago. Since then sediments have been rusted or oxidized. Oxygen is toxic. It rusts a person in a century or less. With oxygen came the danger and blessing of fire. If introduced today, this gas might have difficulty getting approved by the Food and Drug Administration.”

In the intensive care unit, oxygen toxicity during mechanical ventilation is presumed to occur at levels exceeding 0.40. Oxygen is rarely a single factor to produce lung injury. Pulmonary oxygen toxicity is associated with aggressive ventilation, time, and inspired oxygen concentration (FIO₂) [2–5]. Hyperoxia of the lung is presumed to precipitate lung injury through the production of reactive oxygen intermediates [6]. Experimental models of hyperoxic lung injury demonstrate endothelial cell injury, an increase in pulmonary capillary permeability, and a marked increase in inflammatory cells [5].

From a clinician’s perspective, the reduction of FIO₂ to safe levels through the appropriate use of positive end expiratory pressure (PEEP) and manipulation of mean airway pressure is a common and appropriate goal. The ARDSnet ARMA trial utilized a minimum FIO₂ of 0.35 and PEEP of 5 cmH₂O. The PEEP/FIO₂ table was developed through expert opinion with knowledge that treated patients would enter the trial with a PaO₂/FIO₂ <200 mmHg [7]. The more aggressive ARDSnet ALVEOLI trial maintained a lower FIO₂ through manipulation of PEEP, but found no differences in outcomes [8]. Historically, a higher FIO₂ limit often prevents hypoxemia during ICU manipulations and is below the threshold for pulmonary oxygen toxicity.

In this issue of Intensive Care Medicine, de Graaff and colleagues [9] demonstrate that Dutch clinicians appear to be guided by the concept of minimizing FIO₂ to levels that are presumed to be non-toxic with little regard for arterial oxygenation (PaO₂). Their population of predominantly surgical patients (including cardiothoracic and neurosurgical) typically do not require high FIO₂ to maintain arterial normoxemia. The group demonstrated that if FIO₂ was ≤0.40 and PEEP was 5 cmH₂O, even in the presence of hyperoxemia (PaO₂ >120 mmHg) reductions in FIO₂ were only made in 22% of cases. In the presence of hyperoxemia and an FIO₂ >0.40, FIO₂ or PEEP changes were made in 65–82% of cases, with higher FIO₂ resulting in the most frequent changes. These findings suggest that while there is concern about pulmonary toxicity caused by hyperoxic gas mixtures administered to the lungs, there appears to be little concern for hyperoxemia.

The article by de Graaff raises more questions than it answers. Even today, we still do not know what constitutes a safe FIO₂. Data from animal studies and the collective ICU experience over the last 40 years suggest that an FIO₂ <0.40 is non-toxic. The data from de Graaff et al. beg the question, should FIO₂ continue to be decreased in the face of hyperoxemia? Clearly, there is no advantage to oxygen delivery by elevated PaO₂ under normobaric conditions. Exceeding the required FIO₂ to create normoxemia may also mask the presence of worsening pulmonary function [10]. At a PaO₂ of 120 mmHg and oxygen saturation of 99%, a fall in PaO₂ of 50% would still provide an SaO₂ of ≥90%. In this instance, pulse oximetry goals would not be violated, and there would be no warning of a significant decrease in pulmonary dysfunction.

Lung injury occurs over a wide spectrum requiring individualized oxygen requirements and mechanical ventilation needs. Our recent work in closed loop control of oxygenation demonstrates that even severely injured trauma patients (injury severity score >25) can frequently be managed immediately postoperatively with an FIO₂ <0.30. In a number of cases, these patients have had PaO₂ >80 mmHg on an FIO₂ of 0.21 [11]. These data have important implications for military care in austere conditions where oxygen is a limited resource. In contrast, civilian hospitalists rarely worry if their oxygen supply will run out.

A growing body of evidence suggests that hyperoxemia may result in worse outcomes in a number of conditions. Kilgannon et al. reported results of a multi-center trial comparing outcomes in patients after non-traumatic cardiac arrest based on PaO₂. They showed that hyperoxemia (PaO₂ ≥300 mmHg) was associated with increased mortality compared to both the normoxemic (PaO₂ ≤300 mmHg ≥61 mmHg) and the hypoxemic group (PaO₂ ≤60 mmHg) [12]. These findings support earlier animal studies of hyperoxemia following experimental cardiac arrest. In a recent systematic review of oxygen use in the treatment of myocardial infarction, Wijesinghe and colleagues [13] found few studies with significant power to evaluate the safety of oxygen therapy. However, several studies suggest that hyperoxemia may result in vasoconstriction in the coronary circulation and hemodynamic instability [14–17]. These studies do not dissuade the use of oxygen following myocardial infarction, but rather reinforce the need for the titration of oxygen to a normal saturation. The concept of oxygen titration is gaining traction. Recent work with severe traumatic brain injury suggests that both hypoxemia and hyperoxemia are associated with increased mortality and disposition [18]. This is in conflict with the current trend of using hyperoxemia to achieve acceptable levels of brain tissue PO₂ when the deleterious effects of hypoxemia in this population are well described.

The work of de Graaff et al. suggests that FIO₂ titration might best be accomplished using a goal PaO₂/FIO₂ ratio with a lower limit of FIO₂ set at 0.21. It also reminds us that, despite the fact that oxygen is ubiquitous in the hospital and easily delivered, often without a physician order, it is a drug that should be titrated to the desired effect. Both hypoxemia and hyperoxemia have consequences beyond pulmonary oxygen toxicity.

The discovery of oxygen is an interesting account of several investigators arriving at the same conclusion around 1773–1774. Before Lavoisier coined the term ‘oxygen’ it had been called ‘nitro ariel spirits,’ ‘dephlogisticated air,’ ‘fire air,’ and ‘eminently breathable air’ [19]. As oxygen nears its 250th birthday, it is time to treat it as an important drug with consequences for both under and over dosing.