83 4 Comparison of two AOCs in oxygen targeting in preterm infants during admission In this study, where CLiO2 was compared with manual titration, results for hypoxia were similar to our CLiO2 study group, but hyperoxia was slightly lower and time within target range higher. Contrary to the randomised cross-over study14 comparing OxyGenie with CLiO2 for 24-hours periods, OxyGenie control was associated with less time under target range. This may be due to the differences in analysis (in the crossover study periods without supplemental oxygen were included, and intended FiO2 rather than measured FiO2 was collected for both devices). Our choice in definition of room air may have influenced the results. Unfortunately, it was not possible to retrieve the intended FiO2 for the OxyGenie group, as this is not routinely collected, and measured FiO2 was not routinely collected for the CLiO2 group. To increase the validity of the comparison between AVEA and SLE6000, we excluded periods where the inspired oxygen concentration was 23% or lower, reducing the penalty the SLE6000 would inherently receive for measuring more than 21% of inspired oxygen while no supplemental oxygen is given. As such, only less stable episodes with higher risk of hyperoxaemia are studied. Unique in this study is that we investigated the relation between postnatal age and time spent within certain oxygen saturation ranges. As demonstrated in our secondary analysis, comparing achieved target range time can be heavily influenced by the relatively stable respiratory period that usually occurs after a postmenstrual age of 30 weeks. Compared with our primary analysis, one can ascertain all results are diluted by this stable period of near-100% time within target range. Better oxygenation results will not be achieved in phases where infants receive no supplemental oxygen because they are adequately saturated. Indeed, the benefit of an AOC will be modest as then its only role is to respond to intermittent hypoxia, triggered by apnoea or other destabilising events. The focus for research concerning the entire admission should lie on phases where the infant exhibits less respiratory stability and receives supplemental oxygen. Decisions made during the analysis may have influenced the results, but were important to maintain generalisability. Periods in the AVEA epoch where HFO or HFNC was the mode of respiratory support were excluded in the oxygen-only analysis, as there was no automated oxygen control available during these times. Furthermore, we excluded infants requiring less than 72 hours of respiratory support because their data could skew the results while their outcome was based on relatively little information and not representative for the average very preterm infant: they were either too stable and saturated fully immediately or they died soon after birth, which is highly unlikely to be related to choice of oxygen control device.