The dangers of oxygen therapy – hyperoxia and mortality

By measurement of the partial pressure of oxygen in arterial blood (pO₂(a)) and oxygen saturation in arterial blood (sO₂(a)), blood gas analysis provides the means for monitoring supplemental oxygen therapy. Oxygen is prescribed in many medical emergencies in which tissue oxygenation is threatened because of respiratory failure, (defined as pO₂(a) < 8.0 kPa) and/or reduced tissue perfusion. 

Depending on the mode of delivery, the fraction of inspired oxygen (FO₂(I)) associated with oxygen therapy can range from 25 % to 100 %, compared with normal FO₂(I) of 21 % when breathing ambient air at sea level. Since FO₂(I) determines pO₂(a), high-dose oxygen therapy (FO₂(I) > 50 %) can cause pO₂(a) to rise well in excess of the upper limit of the reference range, a condition called hyperoxemia that potentially results in hyperoxia (increased oxygen in tissues). 

Notwithstanding the general appreciation that oxygen in excess is potentially toxic to tissue cells, it has been assumed that, with the notable exception of neonates who are particularly vulnerable, transient hyperoxia is a side effect of high-dose oxygen therapy that is essentially harmless if not unduly prolonged, and well worth the cost of avoiding tissue hypoxia. 

This assumption is now being challenged, and there is a growing body of clinical study directed at establishing the real safety profile of hyperoxia during oxygen therapy. Among them is a recently published study from the Australian and New Zealand Intensive Care Research Centre based at Monash University in Melbourne. 

Investigators here sought to examine the notion, suggested by previous study, that hyperoxia occurring during resuscitation from cardiac arrest is an independent risk factor for death. Clearly if this were indeed the case, more conservative use of supplemental oxygen during cardiac resuscitation would be warranted. 

The Australian researchers used a massive clinical record database relating to all 12,108 patients who had received attempted resuscitation from cardiac arrest at 125 intensive care units across Australia and New Zealand between 2000 and 2009. 

Blood gas results extracted from this database revealed that 1,285 (10.6 %) of these patients had hyperoxia, defined as pO₂(a) > 40.0 kPa (300 mmHg); 8,904 (73.5 %) had either hypoxia, defined as pO₂(a) < 8 kPa (60 mmHg) or reduced oxygen transfer, defined as the ratio of pO₂(a) : FO₂(I) < 40 (< 300 if mmHg is the unit of pO₂(a) measurement); and the remaining 1,919 (15.9 %) patients had normoxia, defined as pO₂(a) in the range of 8.0-40 kPa. 

Overall, 6,968 (58 %) of these cardiac arrest victims did not survive, despite resuscitation measures. Mortality rate was significantly lower (47 %) in the normoxia group than in the hyperoxia group (59 %) and the hypoxia or poor-oxygen-exchange group (60 %). Mortality was, as might be expected, highest (70 %) in the subset of patients in this last group who had hypoxia (pO₂(a) < 8.0 kPa). 

The strength of the apparent association between hyperoxia and increased risk of death was greatly reduced when other factors, (particularly severity of illness) were taken into account. Furthermore, Cox proportional hazards modeling of survival found no independent relationship between the degree of hyperoxia and risk of death. The authors conclude that hyperoxia has “no robust and consistently reproducible independent relationship with mortality”. 

The data suggests that patients with hyperoxia during resuscitation from cardiac arrest are less likely to survive, not because of the hyperoxia per se, but because they are sicker, i.e. already less likely to survive before oxygen administration. Rather than being a contributory cause of death, hyperoxia is more likely just an incidental (innocent) consequence of higher oxygen dose delivered in response to poorer clinical condition.