Automatic versus manual control of oxygen and neonatal clinical outcomes in extremely preterm infants: a multicentre, parallel-group, randomised, controlled, superiority trial.

BACKGROUND: Extremely preterm infants require respiratory support with supplemental oxygen and have frequent hypoxaemic episodes. These episodes, and exposure to inadequately high concentrations of oxygen, are associated with major complications and death. Closed-loop automated control of the fractional concentration of oxygen in inspired air (FiO2-C) reduces the time below and above the target range for the pulse oximeter oxygen saturation (SpO2) and caregivers' workload. We aimed to study whether FiO2-C during respiratory support versus routine manual control might also improve clinical outcomes.

METHODS: This multicentre, parallel-group, randomised, controlled, superiority trial was done in 32 neonatal intensive care units in China, Germany, the Netherlands, and the UK. Infants born at 23+0 weeks to 27+6 weeks postmenstrual age were included and randomly assigned to FiO2-C or routine manual care stratified within centres by postmenstrual age at birth and sex. FiO2-C was provided in addition to routine manual control of FiO2 using infant ventilators. The composite primary endpoint was death, necrotising enterocolitis, or bronchopulmonary dysplasia up to 36 weeks postmenstrual age, or severe retinopathy of prematurity by 44 weeks postmenstrual age. Secondary endpoints were the components of the primary endpoint and the maximum retinopathy of prematurity severity score in either eye on the International Neonatal Consortium Retinopathy of Prematurity Activity Scale. The trial was stopped early because of poor recruitment. The primary analysis included the intention-to-treat population with non-missing primary outcome data. This trial was registered with ClinicalTrials.gov (NCT03168516) and is closed.

FINDINGS: Between July 1, 2018, and Oct 31, 2023, 1082 infants were enrolled and randomly assigned to the FiO2-C group (n=539) or routine manual control group (n=543). Median postmenstrual age was 26+1 weeks (IQR 24+6-27+1). 557 (51%) of 1082 infants were male and 525 (49%) were female. The primary endpoint occurred in 206 (39%) of 534 infants in the FiO2-C group versus 222 (41%) of 538 infants in the routine manual control group (adjusted odds ratio 0·90, 97·5% CI 0·65-1·24; p=0·47). Rates of death (48 [9%] of 536 infants vs 50 [9%] of 541 infants), necrotising enterocolitis (27 [5%] of 538 vs 36 [7%] of 542), bronchopulmonary dysplasia (104 [21%] of 486 vs 110 [23%] of 485), and severe retinopathy of prematurity (86 [18%] of 491 vs 95 [19%] of 496) were also similar between groups. The maximum retinopathy of prematurity severity score was similar between groups (median 7 [IQR 0-9]; p=0·24). Overall, 197 serious adverse events were reported in the FiO2-C group and 192 in the routine manual control group, with no evidence of harm related to the intervention. Postnatal age at death and primary causes of death were similar in both groups. Four serious incidents related to software function without apparent harm to affected infants were reported.

INTERPRETATION: FiO2-C did not improve neonatal clinical outcomes. The long-term application of FiO2-C to reduce caregivers' workload might be considered safe. FiO2-C algorithms should be tested before routine application.

FUNDING: German Federal Ministry of Education and Research.