ObjectiveAutomatic control (SPOC) of the fraction of inspired oxygen (FiO2), based on continuous analysis of pulse oximeter saturation (SpO2), improves the proportion of time preterm infants spend within a specified SpO2-target range (Target%). We evaluated if a revised SPOC algorithm (SPOCnew, including an upper limit for FiO2) compared to both routine manual control (RMC) and the previously tested algorithm (SPOCold, unrestricted maximum FiO2) increases Target%, and evaluated the effect of the pulse oximeter’s averaging time on controlling the SpO2 signal during SPOC periods.DesignUnblinded, randomised controlled crossover study comparing 2 SPOC algorithms and 2 SpO2 averaging times in random order: 12 hours SPOCnew and 12 hours SPOCold (averaging time 2 s or 8 s for 6 hours each) were compared with 6-hour RMC. A generated list of random numbers was used for allocation sequence.SettingUniversity-affiliated tertiary neonatal intensive care unit, GermanyPatientsTwenty-four infants on non-invasive respiratory support with FiO2 >0.21 were analysed (median gestational age at birth, birth weight and age at randomisation were 25.3 weeks, 585 g and 30 days).Main outcome measureTarget%.ResultsMean (SD) [95% CI] Target% was 56% (9) [52, 59] for RMC versus 69% (9) [65, 72] for SPOCold_2s, 70% (7) [67, 73] for SPOCnew_2s, 71% (8) [68, 74] for SPOCold_8s and 72% (8) [69, 75] for SPOCnew_8s.ConclusionsIrrespective of SpO2-averaging time, Target% was higher with both SPOC algorithms compared to RMC. Despite limiting the maximum FiO2, SPOCnew remained significantly better at maintaining SpO2 within target range compared to RMC.Trial registrationNCT03785899