Hylke Salverda

163 9 General discussion and future perspectives General conclusion Automated oxygen control is evolving and the technology holds promise. It increases time spent within the prescribed oxygen saturation target range, decreases hypoxia and hyperoxia, and reduces workload. This thesis is the first to show that these oxygenation outcomes are influenced by choice of automated oxygen controller. The OxyGenie controller was more effective in keeping oxygen saturation within the target range and preventing hyperoxaemia when compared to the CLiO2 controller. We confirmed that these findings also apply during the entire NICU stay. Ultimately, our results demonstrated that OxyGenie is a better choice than CLiO2 for oxygenation targeting. The influence of automated oxygen control on mortality, morbidities and neurodevelopmental outcome at two years corrected age is not yet clear. However, OxyGenie control was associated with better clinical outcome than CLiO2 control, strengthening the suggestion of controller influence on outcome. The retrospective nature of the studies precludes us from drawing definite conclusions on the causal effect of choice of algorithm on outcome, but the data from the studies performed are pointing in the same direction. Future perspectives Several automated oxygen control algorithms are embedded in commercially available ventilators, each of which has its own design and strategy. 15, 16, 48-51 Choice of design will influence how successful oxygen titration will be. As the influence on clinical effect of better titration may be small, the most effective algorithm should be used. Direct head-to-head comparison can be performed using a cross-over design in preterm infants, but changes in respiratory condition would likely preclude testing more than two algorithms at a time. Direct comparison of all algorithms would entail testing them under the same circumstances. A bench test incorporating a model of a preterm infant could be used, against which each oxygen control algorithm could be tested. We are in the process of developing such a model. The strength of this design is that each ventilator would be tested against the same ‘typical’ patients, which is helpful in informing clinicians exactly what to expect when using an algorithm/ ventilator combination. To date it remains unclear what the least harmful range to target is, which could be dependent on what technique is used control the titration of oxygen. As was seen in the NeOProM studies52 and during another study27, during manual titration