(P23-19-26) Understanding the Biological & Social Determinants of the Human Milk Metabolome

Objectives: Human milk (HM) contains metabolites that reflect maternal phenotype and support infant growth, development, and long-term health. Most HM metabolomics studies have focused on small cohorts of healthy, exclusively breastfeeding mothers. Therefore, we investigated determinants of the HM metabolome at 2 weeks postpartum in a large and diverse cohort of mixed-feeding dyads from Cincinnati, OH.

Methods: HM samples (n=1091) were collected at 2 weeks postpartum and analyzed using ¹H-nuclear magnetic resonance spectroscopy. Fifty-seven HM metabolites were quantified and pseudo-log transformed to reduce skew and stabilize variance. Sociodemographic, maternal health, delivery, and infant-related data were self-reported or extracted from medical records. Maternal secretor status was defined by detectable 2'-fucosyllactose. Group least absolute shrinkage and selection operator (LASSO) regression identified predictors of individual HM metabolites. Potential predictors and select interactions were chosen a priori based on prior literature and biological relevance. After removing samples with missing data (n=168), data were split 80/20 into training and testing sets. Models for each HM metabolite were trained using 10-fold cross-validation with the one-standard-error rule to prioritize sparsity and avoid overfitting.

Results: Over 60% (n=35/57) of HM metabolites had predictive models with one or more non-zero coefficients. As expected, maternal secretor status was the dominant predictor for 7 of the 10 human milk oligosaccharides (HMO). Primarily formula feeding (≥ 50% of feeds) was the most influential predictor for 12 of 25 non-HMO metabolites and was associated with higher acetylcarnitine (R² = 0.27) and lower glucose (R² = 0.20), creatine phosphate (R² = 0.18), and phosphocholine (R² = 0.11). Race and income were also influential predictors of the non-HMO metabolites, primarily impacting pantothenate (R² = 0.20) and urea (R² = 0.14).

Conclusions: Maternal secretor status and formula supplementation were the primary determinants of the HM metabolome at 2 weeks postpartum. Future research will investigate whether the differences related to formula feeding reflect low milk supply or breastfeeding practices.

Funding Sources: NICHD, NIAID