Article Text
Statistics from Altmetric.com
Oxygen is one of the drugs most frequently used in neonates, yet often with the highest concentrations given to those with the least developed defence mechanisms to its potentially toxic side effects. To what extent even minor variations in blood oxygen levels may affect longer-term outcomes such as mortality, retinopathy of prematurity (ROP) or necrotising enterocolitis (NEC) was recently shown in the large trials contributing to the NEOPROM collaboration.1
In addition to the importance of closely maintaining an assigned baseline target range, avoiding intermittent hypoxaemia or hyperoxaemia (eg, a pulse oximeter saturation (SpO2)<80% or >95%) may be equally important.2 Given this situation, combined with the high workload the nursing staff caring for preterm infants often face, close control of oxygen levels is a task that may best be left to a computer algorithm rather than the bedside nurse.
The properties required of such algorithms, however, are manifold: they should be able to respond to both, a gradual change in oxygen requirements and sudden hypoxaemia, and should also be capable of avoiding the build-up of increasing fluctuations in FiO2/SpO2 during periods with oscillating SpO2 values (eg, during periodic breathing). Given these requirements, it may not come as a surprise that studies validating such devices are only slowly becoming available, while comparative data on their effectiveness are yet lacking altogether.
Most FiO2 controllers currently available are rule-based, that is, respond to any deviation from target SpO2 according to a set of predefined rules, while others are more flexible, that is, try to adapt their response to variations in an infant's lung function over time (eg, proportional-integral-derivative (PID) controllers).
A group from Tasmania reports on their contribution to this field.3 ,4 They developed an FiO2 controller based on a PID algorithm that was designed to avoid hyperoxaemia and adapt to the presence of severe V/Q mismatch. They first tested their device in a simulator that was fed with FiO2 and SpO2 data recorded in preterm infants receiving continuous positive airway pressure support and compared its simulated performance with (i) a controller that only responded with a 30 s delay executed after each change in FiO2 and (ii) a controller that did not have an integral term, that is, operated more like a rule-based controller. This set-up allowed them to enhance their controller's performance, which was then tested in real patients on their neonatal intensive care unit. Here, the authors performed three 4-hour serial measurement periods of SpO2, first during routine manual control, then under automated control and finally again during routine manual control. Compared with routine nursing care, their controller improved the proportion of time (%time) spent in target (91%–95% SpO2) from 55% to 78%, and reduced the %time spent at <80% SpO2 from a median of 0.7% to 0.0% and that in hyperoxaemia (SpO2 97%–99%) from 5.0% to 0.7%.
Two questions arise from these clinical data: (i) how do they compare with those reported for other (possibly less sophisticated) control algorithms, and (ii) does the rather narrow target range used here make it easier or more difficult for a controller or a nurse to maintain an infant's SpO2 values within this range?
Plotting the data on %time within target from available cross-over studies,4–11 it first becomes clear that in each study (or subgroup) the automated FiO2 controller was better at keeping infants within their assigned range than the routine beside nurse was (figure 1). The %time spent in target, however, varied widely between studies, from 32% to 82% for manual control and from 40% to 91% for automated control. The performance of both the controller and the nursing staff did not appear to be related to the width of the target range (figure 1) or the underlying variability in SpO2 levels during manual control: In one of the above studies, performed in four centres, two used similar target ranges (83%–93% and 85%–94% SpO2, respectively), but one had an increase in %time in target from 43% to 66%, while the other improved it from 65% to 84%, that is, in both cases by approximately 20%, but from rather different baselines achieved during routine nursing care.
Independent of such performance characteristics, recent studies raised concern that any effect automated FiO2 control may have on clinical outcomes may be smaller than we would hope. This is because the association between intermittent hypoxaemia and longer-term outcomes seems to be stronger for episodes occurring at 6–10 weeks of age than in the first few postnatal weeks, suggesting that the brain may become more sensitive to the deleterious effects of intermittent hypoxaemia as it becomes more mature.2 Even assuming the associations found reflect causality, automated FiO2 control is not very likely to ameliorate this as most infants will already be in room air or on very little respiratory support by 6–10 weeks, that is, not on an FiO2 controller anymore.
Also, a recent study examining the effects of automated FiO2 control on measures of cerebral oxygenation determined by near-infrared spectroscopy did not indicate that a higher %time in target measured by pulse oximetry translated into enhanced stability of cerebral oxygenation.12 Although preliminary, such findings raise doubts whether improved control of peripheral SpO2 will ultimately improve clinically relevant outcomes.
Thus, as much as an improvement by 23% for %time in target, as now found by Plottier et al, is impressive, it does not appear to be qualitatively different from results found for rule-based controllers, and it is yet unknown whether the introduction of this new technology will help not only to facilitate nursing care, but ultimately improve patient outcomes. Here, it is of utmost importance to address this question in a rigorously designed multicentre study before these devices are becoming standard of care. In this regard, it is most welcome that the German Ministry of Research and Education recently granted the authors funding for a large multicentre parallel group trial to test whether automated FiO2 control, compared with standard manual adjustments, will result in better patient outcomes, that is, fewer deaths or severe ROP, bronchopulmonary dysplasia or NEC.
Such data on clinically relevant outcomes are needed before this new technology becomes part of routine neonatal care. In addition to testing whether automated FiO2 control may result in less oxygen-related complications, such a study should also try to address whether one controller performs better than another, and perhaps also whether this will translate into differences in clinical outcomes. Fortunately, as funding for performing such a study has now been secured, it looks as if we will know more on this topic in a few years time.
References
Footnotes
Contributors CFP wrote the first draft of this manuscript, ARF critically reviewed it and added significant intellectual content.
Competing interests None declared.
Provenance and peer review Commissioned; internally peer reviewed.
Linked Articles
- Original article
- Original article