Single-center, randomised controlled, cross-over clinical trial in preterm infants born at gestational age below 34+1/7 weeks receiving supplemental oxygen and respiratory support (continous positive airway pressure (CPAP) or non-invasive ventilation (NIV) or invasive ventilation (IV)). Routine manual control (RMC) of the fraction of inspired oxygen (FiO2) will be tested against RMC supported by automatic control (SPOC) with "old"-algorithm and RMC supported by CLAC with "new"-algorithm.

The first primary hypothesis is, that the use of the "new" algorithm results in more time within arterial oxygen saturation (SpO2) target range compared to RMC only. The a-priori subordinate hypothesis is, that the new algorithm results in more time within SpO2 target range compared to SPOCold.

The second primary hypothesis is, that the use of 2 seconds averaging time of the SpO2 Signal results in more time within arterial oxygen saturation (SpO2) target range compared to the use of 8 seconds averaging interval of the SpO2 signal.

BACKGROUND AND OBJECTIVE In preterm infants receiving supplemental oxygen, routine manual control (RMC) of the fraction of inspired oxygen (FiO2) is often difficult and time consuming. The investigators developed a system for closed-loop automatic control (SPOC) of the FiO2. The objective of this study is to test a revised, "new" algorithm with 3 adaptions against the former "old" algorithm and against RMC. The 3 adaptions are:

1. Faster re-adjustment to baseline-FiO2 (baseline FiO2: mean FiO2 during the previous 5min)
2. Delayed reduction of FiO2 below baseline FiO2
3. Maximum FiO2 adjustable by user

The first primary hypothesis is, that the application of SPOCnew in addition to RMC results in more time within arterial oxygen saturation (SpO2) target range compared to RMC only. The a-priori subordinate hypothesis is, that the revised algorithm is more effective as the old algorithm to maintain the SpO2 in the target range.

The second primary hypothesis is, that the shortening of averaging time used for the SpO2 Signal from 8 seconds to 2 seconds results in more time within SpO2 target range for both, SPOCnew and SPOCold.

Further hypotheses for exploratory testing are, that the SPOC new algorithm will achieve a lower proportion of time with SpO2 above and below the target range, hyper- and hypoxia and an improved stability of cerebral oxygenation (measured as rcStO2 and rcFtO2E determined by Near-infrared spectroscopy) compared with SPOCold and RMC. Reduction of staff workload (estimated by number of manual adjustments per hour) by SPOC. Validation of a clinical scoring tool to monitor severity of apnea of prematurity.

STUDY DESIGN The Study is designed as a single-center, randomized controlled, cross-over clinical trial in preterm infants receiving mechanical ventilation or nasal continuous positive airway pressure or non-invasive ventilation and supplemental oxygen (FiO2 above 0.21). Within a 30-hour period the investigators will compare 6 hours of RMC with 12-hour periods of RMC supported by SPOCnew algorithm or SPOCold algorithm, respectively. During intervals with SPOC control the SpO2 Signal averaging time will be 2 second or 8seconds , respectively, for 6 hours each.

Inclusion Criteria:

• gestational age at birth <34+1/7weeks and
• invasive mechanical ventilation OR noninvasive ventilation OR continous positive airway pressure support and
• Fraction of inspired oxygen above 0.21 before inclusion and
• more than 2 hypoxaemic events (arterial oxygen saturation below 80%) within 8 hours before inclusion and
• parental written informed consent

Exclusion Criteria (any of the following):

• congenital pulmonary anomalies
• congenital heart defects influencing SpO2 (i.e. cyanotic heart defects)
• right-to -left shunt through a PDA
• Severe neonatal complications during study period (sepsis, necrotising enterocolitis)
• diaphragmatic hernia or other diaphragmatic disorders