Skip to main content
NUTRITION 2026 Main banner

(PTFS03-02-26) Early Gut Microbes Are Important Predictors of Future Weight in Breastfed Infants

Saturday, July 25, 2026
1:04 PM - 1:08 PM ET
Location: Prince George's Hall B-E (The Hub) - Poster Theater 2 / Poster Board Number: PT2-2
Early Career Research Award Competition
Disclosure(s):
Andrew Oliver, PhD: No relevant financial relationship(s) with ineligible companies to disclose.
Daniela Hampel, PhD: No relevant financial relationship(s) with ineligible companies to disclose.
Danielle G. Lemay, PhD: No relevant financial relationship(s) with ineligible companies to disclose.
Objectives: Human milk can provide infants with complete nutrition in the first six months. However, some energy sources found in milk are only digestible by specific microorganisms in the infant gut. Whether these microbes influence infant growth is unknown. The objective of this study was to determine if gut microbes contribute to the growth of infants born to well-nourished mothers. We hypothesized that models of infant growth would improve with the addition of microbial features. We also predicted there may be a “lag” effect: microbial abundances from earlier visits may better predict infant weight than microbial abundances measured at the same visit.

Methods: We leveraged the Mothers, Infants, and Lactation Quality (MILQ) cohort, which enrolled participants from Denmark, Brazil, Bangladesh, and The Gambia. In a subset of MILQ infants (n=560), we sequenced shotgun metagenomes from visits M1 [1–3.49 mo], M2 [3.5–5.99 mo], and M3 [6–8.5 mo]. Using random forest regression, we predicted weight-for-age z-scores (WAZ) given baseline (e.g., birth weight, maternal anthropometrics, etc.) + infant diet predictors (milk volume, milk macronutrients, and non-milk intake) (B+ID) or baseline + infant diet + microbiome (B+ID+M).

Results: Averaged across 50 random seeds, the best B+ID models predicting WAZ occurred at M1 (R2 of 0.511, 95% CI [0.49, 0.53]). Using SHAP analysis, top features driving the models routinely included birth weight and milk volume consumed. For visits M2 and M3, B+ID+M models predicted WAZ significantly better than B+ID models (pairwise Wilcoxon, p.adj < 0.05). We also found the microbiome may have a “lag” effect on WAZ; WAZ at M2 was significantly better predicted using the microbiome at M1 (R2 of 0.484, 95% CI [0.45, 0.52]) than the microbiome at M2 (R2 of 0.378, 95% CI [0.35, 0.41]) (p.adj < 0.001). The abundance of Bifidobacterium breve was frequently an important feature as were microbial genes, such as beta-galactosidase. Beta-galactosidase abundance was significantly correlated with WAZ at 1-3.5 months (partial correlation, r = 0.14, p < 0.05).

Conclusions: An infant’s birthweight and the volume of milk are the strongest determinants of the infant’s WAZ. However, adding microbiome information tends to increase model performance, especially if the microbiome was measured prior to weight assessment.

Funding Sources: Gates Foundation