42 Chapter 2 strategy for oxygen titration. It is possible that some of these improvements have already been included in contemporary algorithms, but unfortunately details on the exact operation of most algorithms are lacking in literature. It should be noted that extra input also gives a potential source of error. Although AOC is more effective in keeping SpO2 in TR when compared to manual control, the effect on clinical outcome is currently unknown. Studies are needed using a large sample size where continuous AOC is used for a longer period during NICU admission. Currently the FiO2-C study is aiming to provide such information by recruiting infants < 28 weeks of gestation, with all available control algorithms allowed to be used.47. All studies of AOC to date report an increase in time spent in TR, with less hyperoxia and severe hypoxia, both in frequency and duration. Also, there is an abundant evidence underlining the deleterious effect of hypoxia and hyperoxia3-5, 48-51. It is therefore reasonable to expect that these devices will change outcomes. Recent trials comparing a low versus a high oxygen saturation target once again emphasised the importance of which oxygen saturation to target.52 The masking of clinical deterioration is an often-mentioned concern of AOC, which could lead to later detection than when nurses handle oxygen manually. However, this concern can be countered by alarming clinicians when the baseline oxygen requirement is increased above a certain threshold. Since the study by van Zanten et al.33, implementation of the CLiO 2-algorithm as part of standard care in the NICU of Leiden has been successful. Three years after implementation caregivers are experienced in using and interpreting the handling of the controller. BaseFiO2 and the amount and magnitude of interventions performed by AOC are assessed regularly, now providing us with a novel indicator of clinical deterioration. Therefore, to utilise the full potential and to safely apply AOC, adequate training should be a prerequisite.