To put it bluntly there has never been a shred of convincing evidence to guide limits for the rational use of supplemental oxygen in the care of extremely premature infants. For decades, the optimum range of oxygenation . . . was, and remains to this day, unknown.
WA Silverman 20041
Attempted revival of apparently lifeless newborns and death from respiratory failure among preterm infants was reported in ancient times.2 Oxygen was described by Priestly and Lavoisier in the 1770s. Illustrating the enduring enthusiasm of physicians for introducing new untested therapies, oxygen was given to revive newborns in 1780, within 6 years of its discovery.2 Positive pressure ventilation went in and out of fashion for reviving newborns over much of the following 200 years. Oxygen was also intermittently given, either during positive pressure ventilation, via hyperbaric chambers3,4 or directly into the stomach.5 Adverse effects of oxygen were demonstrated in the 1950s by the epidemic of blindness due to retinopathy of prematurity in premature infants nursed in high levels of oxygen to regularise periodic breathing.1
Animal experiments performed by Geoffrey Dawes’ group in the 1950–1960s underpin neonatal resuscitation as we know it today.6 In the seminal experiment a term Rhesus monkey was acutely and profoundly asphyxiated by having its umbilical cord ligated and its head covered with a saline-filled condom to prevent breathing. Initial gasping respiration was followed by apnoea, then further agonal gasps before breathing ceased completely. This was accompanied by falling heart rate and blood pressure, worsening acidaemia and inexorable progression towards death unless the infant was revived by positive pressure ventilation using an oxygen-enriched gas. Organised resuscitation courses evolved over the decades that followed. Based on a consensus of US opinion leaders as to what was ‘accepted’ practice, the first textbook of the Neonatal Resuscitation Program (NRP) was published in 1987.7 Latterly the Neonatal Subcommittee of the International Liaison Committee on Resuscitation (ILCOR) has made recommendations for practice8 and reconvened at five yearly intervals to update them.9,10 It is abundantly clear from this process of evidence evaluation that there is a paucity of good-quality evidence to support current practice.
In contrast to many other aspects of neonatal resuscitation, the role of oxygen has been extensively studied in recent years. Until comparatively recently, 100% oxygen was recommended for all newborns who were judged to need any support at birth.8 It is easy to understand how this may have evolved and appeared reasonable. Compared to humans at other ages, newborns have low oxygen saturation of haemoglobin in their arterial blood and, hence, are cyanotic.11,12 Though this is the state of nature, it has probably made many who have observed it uncomfortable and encouraged them to give oxygen. Concerns were raised that the use high concentrations of oxygen in infants could lead to the formation of reactive oxygen species which have deleterious effects on cells and may exacerbate reperfusion injury. Much basic science and animal work ensued which suggested 100% oxygen might be harmful. Several randomised trials comparing air to 100% oxygen for the resuscitation of asphyxiated term or near-term human infants followed.13–16 The largest of these trials had methodological limitations and none had detailed long-term neurodevelopmental follow-up.12,13,16 Uncomfortably, the results of meta-analysis showed a reduction in neonatal mortality for infants randomly assigned to resuscitation with air.17 Adverting to the limitations of the included studies, ILCOR stated in 2006 that there was insufficient evidence on which to base firm recommendations for the optimal oxygen concentration to use; and that preterm infants were likely to be more susceptible to injury from oxygen due to their underdeveloped anti-oxidant systems.10 Based on this statement, the NRP recommended using 100% oxygen for term infants and lower oxygen concentrations guided by pulse oximetry for preterm infants.7 Bearing in mind the tenet of evidence-based medicine that practice should be based on the best evidence available, these recommendations seem somewhat incongruous. At that time, the best (imperfect) evidence suggested air was preferable for term infants; and studies of preterm infants were conspicuous by their absence, a point specifically noted by ILCOR.10 It is, however, probably unfair to criticise for this paradox in isolation, considering the model on which our practice is based is an acutely asphyxiated term monkey (whose relevance to preterm infants, who are infrequently acutely asphyxiated, is at best dubious).
In this edition of the journal, Clark's survey18 shows variation in oxygen supplementation of preterm infants at birth in tertiary neonatal centres in Australia and New Zealand in 2006. It demonstrates change over recent years – in 2004 oxygen blenders were available in delivery rooms in 36% of centres;19 in 2006, 64% of centres had blenders.18 Clark's study also shows considerable variation in the attitudes of nurses and doctors to oxygen for preterm infants at birth. This variation reflects uncertainty as to how best to manage these infants. This may seem unsettling or disappointing; on the contrary I find it reassuring. It illustrates a growing awareness that we may do harm with oxygen. Until comparatively recently, there was consensus that all infants should receive 100% oxygen. As many a lemming could attest, consensus among peers is not necessarily a good thing. To quote Silverman, ‘it is encouraging to see that neonatal medicine is beginning to wake up after years of dogmatic slumber.’1
Since the last iteration of the ILCOR guidelines, more studies and meta-analyses have been published that support the use of air for term infants.20–22 Further studies show that oximetry in the delivery room is feasible23 and describe the changes in oxygen saturations of healthy infants post-natally.24–26 More recently, small randomised trials and cohort studies comparing resuscitation of extremely preterm infants with 100% oxygen to variable concentrations of oxygen guided by pulse oximetry have appeared.27–30 These studies measured short-term outcomes, did not report long-term follow-up and were not powered to detect differences in important clinical outcomes. These studies have shown that while supplemental oxygen is often required to achieve median oxygen saturation values seen in healthy term infants in the first minutes of life, usually a lot less than 100% is required and that much less oxygen can be given overall.
Evolution appears to have decided that the optimal concentration of oxygen in inspired gas for mammals is 21%. Somewhat perversely, because the introduction of 100% oxygen pre-dated the exacting standards expected for a change in practice today, the burden of proof lies on air. I guess that for most extremely preterm infants with (and particularly those without) parenchymal lung disease, the optimal concentration lies closer to 21 than 100%. It is probably naïve to think that ‘one size fits all’, and biologic variability among these infants makes it likely that they will vary in how much oxygen they need. Until trials examining variable oxygen concentrations which are adequately powered to detect differences in important clinical outcomes are completed and reported, uncertainty will persist. Whether or not this uncertainty makes us uncomfortable may be interesting; it is, however, largely irrelevant. How we deal with this uncertainty – either by bumbling along or by trying to resolve it in a rational fashion – is far more important. Studies of variable oxygen targeted to oximetry are being planned; they deserve support.