Mechanisms affecting the gut of preterm infants in enteral feeding trials: a nested cohort within a randomised controlled trial of lactoferrin

**Background:** Late-onset sepsis (LOS) and necrotising enterocolitis (NEC) affect 25–50% and 5–7% of very preterm infants respectively, contributing to 20–30% of preterm deaths. Mother's own milk is protective, but mechanisms remain poorly understood. Lactoferrin, an abundant milk glycoprotein with antimicrobial and immunomodulatory properties, was hypothesized to reduce LOS and NEC by modulating the gut microbiome. Meta-analyses of early trials suggested benefit (NNT 17 for LOS prevention), but the large ELFIN RCT (n=2203) found no clinical benefit. The MAGPIE study was nested within ELFIN to investigate mechanisms.

**Methods:** 479 infants <32 weeks' gestation were recruited from 13 UK NICUs between June 2016 and September 2017, with 467 providing usable samples (10,990 stool and 22,341 urine samples). Microbiome analysis used 16S rRNA gene sequencing on 1,304 stools from 201 infants. LC-MS metabolomics was performed on 171 stools (83 infants) and 225 urines (90 infants). GC-MS VOC analysis was performed on 314 stools (117 infants). Infants were matched across trial arms for site, gestation, and clinical variables. Statistical models included gestational age, birth weight, milk type, health status, IMP receipt, day of life, and NICU site as fixed effects, with infant identity as random effect.

**Key Results:** Gut microbiome Shannon diversity at 34 weeks corrected age was not significantly different between lactoferrin (mean=1.24) and placebo (mean=1.06) groups (p=0.11). Lactoferrin receipt explained less than 1% variance in microbiome compositions between groups and a mean of 3% variance across all timepoints and omics layers. NICU site had the greatest influence, explaining a mean 30% variance across all timepoints (up to 40% for VOC composition). Postnatal age explained a mean 7% variance. From DOL 7, NICU site explained 16.7% of bacterial community variance (p<0.01); before DOL 7, milk type had the greatest impact (3.5% variance, p=0.03). Across all NICUs combined, only three unidentified stool metabolites and acetic acid were significantly higher with lactoferrin, while one unidentified metabolite and 2-methyl-propanal were higher with placebo. No bacterial genera remained significantly associated with either trial arm in combined analysis. Within individual NICUs, 19 bacterial genera and 1,664 stool/321 urine metabolite features differed between arms, but few were consistent across sites. Lactobacillus was significantly reduced in lactoferrin recipients at probiotic-administering sites (p<0.02).

**Clinical Implications:** This large multiomic mechanistic study nested within an RCT demonstrates that supplemental bovine lactoferrin has minimal impact on the preterm gut microbiome and metabolome, consistent with the null clinical findings of the parent ELFIN trial. The dominant influence of NICU site—likely reflecting differences in antibiotic use, probiotic administration, and feeding practices—highlights the importance of multisite designs and the limitations of single-center microbiome studies. The findings suggest that single-nutrient supplementation may be insufficient to alter the gut ecosystem in preterm infants, potentially because lactoferrin requires cofactors (e.g., osteopontin) for optimal bioactivity, or because the intervention window is too narrow (median enrolment at day 4, with LOS onset at day 12 and NEC at day 17). The study demonstrates the feasibility and value of nesting mechanistic research within RCTs to inform trial interpretation and future intervention design.