Design: stepped-wedge cluster-randomised hybrid effectiveness-implementation trial in 15 level I NICUs
Population: 1,627 very low birth weight infants (831 intervention, 796 control), Germany, April 2022 to March 2024
Primary outcome: exclusive mother's own milk feeding at NICU discharge
Effect: +11 percentage points (95% CI 3.0–19.0, p = 0.007) from a 43% baseline; no difference in complications or length of stay

Neo-MILK was a stepped-wedge cluster-randomised trial across 15 German level I neonatal units between April 2022 and March 2024, enrolling 1,627 very low birth weight infants — 831 in intervention periods, 796 in control. The intervention was not a single thing: structured lactation counselling, multilingual materials, a mobile app for parents, and the establishment of human donor milk banking, supported by video-based staff training, written guidelines and commitment prompts for medical staff.

Exclusive mother's own milk feeding at discharge rose by 11 percentage points (95% CI 3.0 to 19.0, p = 0.007) from a baseline of 43%, with exclusive formula feeding falling correspondingly. Clinical complications and length of stay did not differ. Prespecified subgroups showed no significant effect in infants under 1,000 g or in those with a CRIB score of 11 or more — the sickest infants, where a mother's supply is hardest to establish and where the intervention did least.

Implementation data are reported alongside, which is unusual and useful. Acceptability, appropriateness and feasibility were high; antenatal counselling reached 71% of intended cases; donor milk banks were established at only 40% of sites; and 61% of staff reported the changes had become routine. Economic modelling suggested net savings nationally. The honest summary is that a bundle of ordinary organisational measures moved a hard outcome by a real margin, and that the gain sat with the less sick infants.

Audit your unit's exclusive mother's milk rate at discharge as a baseline before changing anything
Start counselling antenatally where preterm delivery is anticipated, not after admission
Do not expect the same gain in extremely low birth weight or high-CRIB infants
Treat donor milk banking as the hardest component to implement — only 40% of sites managed it

Why it matters

Feeding outcomes in neonatal units are usually treated as a matter of individual counselling rather than unit design.

The statistics, in plain English

An 11 percentage point rise from a 43% baseline is a quarter more infants going home on mother's milk alone. The interval of 3.0 to 19.0 is wide, so the true gain could be modest or large, but it excludes zero. The stepped-wedge design means all units eventually received the intervention, which controls for unit differences but is vulnerable to secular trends — anything else improving over the two years would inflate the estimate. The null subgroup findings are underpowered rather than proof of no effect.