Hylke Salverda

40 Chapter 2 magnitude, pushing FiO2 up. To prevent inappropriately high oxygen administration, the set FiO2 is capped to 40% above the reference FiO2. The algorithm adjusts for the non-linearity of the oxygen-dissociation curve by applying an error multiplier to positive errors while the integral term remains positive. This rapidly decreases the integral term towards zero, leading to FiO2 reduction and hopefully eliminating overshoot. Whilst in 21% oxygen, values of SpO2 above TR are not considered to represent hyperoxia and the integral term is therefore not altered. Lastly, the handling of the derivative term is modified during hyperoxia. A negative SpO2 slope is nullified if all last five SpO2 values are above the TR midpoint. This means that during hyperoxia a negative slope will not drive up the FiO2. 21, 36, 41 Clinical effect A forerunner algorithm, VDL 1.0 was first tested and refined in a (validated) computer model for the respiratory system of a preterm infant using paired values for ventilation-perfusion (V/Q) ratio and shunt (Qs/Qt) (based on 3788 hours of clinical data).22 Consequently a study36 involving 20 preterm infants on non-invasive respiratory support and supplemental oxygen was done, in which a 4-hour period of AOC was compared with manual control during flanking 4-hour periods (Table 1). Time spent within TR increased during AOC, whereas hypoxia with SpO2 <80% and hyperoxia with SpO2 >98% greatly reduced (0.7% (0.1%-1.3%) vs 0% (0%-0.17%); p=0.0006 and 0.46% (0.22%-1.4%) vs 0% (0%-0.12%); p=0.0010, respectively).36 Discussion Although all contemporary oxygen controlling algorithms appear to increase time spent within TR, the most effective strategy is currently unknown. Comparing these algorithms in the available studies is hindered by variation in TR, pulse oximeter settings, ventilator mechanics, patient populations, modes of respiratory support and aims between studies. A direct head-to-head comparison of algorithms could be performed in infants using a cross-over design, but changes in respiratory condition would likely preclude testing of more than two algorithms at a time. Ultimately the most meaningful and informative comparison of all algorithms may therefore be achieved by bench-top testing of the algorithms embedded in their different ventilators using in silico patient simulations. Titration of FiO2 is challenging for caregivers, especially during hypoxic and bradycardic events related to apnoea of prematurity.39 In addition to better TR adherence, Claure et al. demonstrated that AOC leads to a significant reduction in