The study, conducted by the Neo-MILK Consortium and published in BMC Medicine, was a stepped-wedge cluster-randomized trial carried out in fifteen German Level I neonatal intensive care units between April 2022 and March 2024. In a stepped-wedge design, all participating units eventually receive the intervention, but the timing of the switch is randomized, so that at any given moment some units are providing the new program while others are still delivering standard care. This approach is particularly well suited to hospital-level interventions that cannot easily be withheld from individual patients, and it allowed the researchers to compare outcomes before and after implementation within the same institutions while controlling for temporal trends. The trial was registered in the German Clinical Trials Register under DRKS00025058 and was designed as a hybrid type 1 effectiveness-implementation study, meaning that its primary aim was to test whether the intervention worked, with a secondary focus on how well it could be put into practice.

The intervention itself, known as Neo-MILK, combined two complementary strategies. The first was structured lactation support: standardized counselling for mothers beginning before delivery where possible, multilingual written and digital materials, and a mobile application that gave parents guidance and encouragement throughout what is often an exhausting and emotionally fraught hospital stay. The second component was the establishment of human donor milk banking, which allows pasteurized milk from screened donor mothers to be given to infants whose own mothers cannot yet supply enough milk. Alongside these clinical elements, the researchers deployed a set of implementation strategies aimed at the hospitals themselves, including video-based staff training, written guidelines, commitment nudges designed to prompt medical staff to prioritize human milk feeding, and practical assistance in setting up donor milk banks where none existed.

The primary outcome was the proportion of very low birth weight infants discharged home fed exclusively on their mother’s own milk. Among the 1,627 infants enrolled, 831 in the intervention phase and 796 in the control phase, the intervention raised exclusive mother’s own milk feeding at discharge by 11 percentage points, from a baseline of 43 percent, with a 95 percent confidence interval of 3.0 to 19.0 percentage points and a p-value of 0.007. In parallel, the proportion of infants fed exclusively on formula declined correspondingly, suggesting that the program did not merely reshuffle feeding categories but genuinely converted formula-dependent infants to human milk feeding. For a population in which every milliliter of mother’s milk is associated with reduced risks of necrotizing enterocolitis, infections, and other life-threatening complications, an eleven-point gain achieved through organizational change rather than new drugs or devices is a striking result.

Equally important is what the trial did not find. Rates of clinical complications and the length of hospital stay showed no differences between the intervention and control periods, indicating that the push toward human milk feeding did not come with unintended harms or delayed discharges. The researchers also prespecified subgroup analyses to determine whether the intervention worked equally well across the spectrum of vulnerability, comparing infants with birth weights below 1,000 grams against those at or above that threshold, and infants with high illness severity, defined by a Clinical Risk Index for Babies score of 11 or greater, against those with lower scores. Here the results were more sobering: no significant effects emerged in extremely low birth weight infants or in the highest-risk group, a pattern that the authors interpret as a signal that the most fragile babies may need additional, tailored strategies beyond the standard support package.

Because the trial was designed as a hybrid effectiveness-implementation study, the team went well beyond the primary clinical endpoint and systematically measured how the intervention was received and sustained. Acceptability, appropriateness, and feasibility were all rated highly by participating units, suggesting that the program fit reasonably well into existing workflows. Fidelity was sufficient overall, although not uniform: prepartal counselling, the earliest and arguably most critical touchpoint, was delivered as intended in 71 percent of cases, leaving room for improvement in reaching mothers before birth. The establishment of human donor milk banks proved to be the most variable element, with banks successfully created at only 40 percent of the participating sites, a finding that underscores how much organizational, regulatory, and logistical work stands behind what sounds like a simple idea of sharing breast milk. On the positive side, sustainability indicators were promising, with 61 percent of staff reporting that the new practices had been integrated into routine care by the end of the study.

The economic evaluation added a dimension that is often missing from neonatal nutrition research. Using time-driven activity-based costing, a method that maps every staff activity and resource involved in delivering care, combined with a budget impact analysis, the researchers estimated the financial consequences of rolling out the program nationally. In the base-case scenario, the analysis suggested potential net savings, meaning that the costs of counselling, training, and donor milk banking could be offset by reductions elsewhere in the care pathway. While the authors are careful to frame this as a scenario-dependent projection rather than a guarantee, the finding matters for health systems weighing whether to fund such programs at scale, particularly in Germany where statutory health insurers collaborated in the study and where the German Innovation Fund of the Federal Joint Committee financed the work under grant 01NVF19027.

The trial’s design deserves attention from anyone following the methodology of complex health interventions. By randomizing the order in which the fifteen units crossed over to the intervention, the stepped-wedge approach preserved the ethical appeal of eventually offering the program to everyone while still generating a controlled comparison. The intention-to-treat analysis framework, the use of intracluster correlation coefficients to account for the clustering of outcomes within hospitals, and the prespecification of subgroups all reflect a rigor that strengthens confidence in the headline finding. At the same time, the variability in donor milk bank uptake across sites illustrates a persistent challenge in implementation science: an intervention is only as strong as the organizational soil in which it is planted, and units differ widely in staffing, culture, and infrastructure.

For clinicians and policymakers, the message of the Neo-MILK trial is twofold. First, structured lactation support combined with donor milk banking is an effective, safe, and potentially cost-saving way to increase exclusive mother’s own milk feeding at discharge for very low birth weight infants, moving a substantial fraction of babies from formula to human milk without increasing complications or length of stay. Second, the work is not finished: the absence of measurable benefit in the smallest and sickest infants, the incomplete delivery of prepartal counselling, and the uneven establishment of donor milk banks all point to the need for ongoing organizational support and strategies tailored to the highest-risk families. As neonatal units around the world grapple with how to raise human milk feeding rates, this German trial provides both a tested blueprint and an honest account of where that blueprint still falls short.

Subject of Research: A stepped-wedge cluster-randomized trial evaluating structured lactation support and human donor milk banking in German neonatal intensive care units

Article Title: Effectiveness of structured lactation support and human donor milk banking in German NICUs: a stepped-wedge cluster-randomized trial

Article References: Köberlein-Neu, J., Dymek, N. G., Zimmer, V., Bommhardt, T., Stirner, A. K., Güldenring, I., Ohnhäuser, T., Wiesen, D., Dresbach, T., Scholten, N., & Neo-MILK Consortium/collaborators (2026). Effectiveness of structured lactation support and human donor milk banking in German NICUs: a stepped-wedge cluster-randomized trial. BMC Medicine. https://doi.org/10.1186/s12916-026-05211-1

Image Credits: AI Generated

DOI: 10.1186/s12916-026-05211-1

Keywords: lactation support, human donor milk, very low birth weight infants, neonatal intensive care, stepped-wedge trial, mother's own milk feeding, donor milk banking, neonatology, implementation science, health economics, BMC Medicine, Neo-MILK