Background: A current standard of care based on expert opinion is to commence parenteral nutrition (PN) within hours of birth in very preterm infants. Trials in critically ill adults and children, including term infants have found short and long-term harms from early initiation of PN.

Methods: We included all infants born below 31 weeks gestation between January 2008 and December 2019 and admitted to National Health Service neonatal units in England and Wales. The source of data was the National Neonatal Research Database. The exposure was PN initiated within the first two days after birth (early) versus after the second postnatal day (late). We used propensity matched analysis to balance the two groups on background variables. The primary outcome was survival to discharge without major morbidity.

Findings: Of 65,033 infants included, 16,294 infants formed the matched cohort, 8147 in each group. There was no evidence of a difference in survival to discharge without major morbidity (absolute rate difference (ARD) between early versus late 0·50%; 95% Confidence Interval (CI), -1·45, 0·45; p=0·29). Survival to discharge was higher in the early group (ARD -3·25%; 95% CI, -3·82 to -2·68; p<0·001) but they also had higher rates of late-onset sepsis (ARD -0·84%; 95% CI, -1·20 to -0·48; p<0·001), bronchopulmonary dysplasia (ARD -1·24%; 95% CI, -2·17 to -0·30; p=0·01), treatment for retinopathy of prematurity (ARD (-0·5%; 95% CI, -0·84 to -0.17; p<0·001), surgical procedures (ARD -0·8%; 95% CI, -1·40 to -0·20; p=0·01) and greater drop in weight z-score between birth and discharge (absolute difference 0·019; 95% CI, 0·003 to 0·039; p=0·02). Among infants that died the median age (days) at death was shorter in the late group (ARD 6; 95% CI, 6; p<0.001).

Interpretation: These observational data justify the need for a randomised controlled trial of early versus late PN powered to detect differences in long-term functional outcomes.