Gut Microbiota in the First 2 Years of Life and the Association with Body Mass Index at Age 12 in a Norwegian Birth Cohort

Amount of variation in BMI z-score at age 12 explained by the infant gut microbiota at each sampling time. This plot shows the estimated R² values with 95% confidence intervals from the repeated cross-validation of the random forests to predict childhood BMI z-scores based on infant gut microbiota characteristics at each sampling time, both unadjusted and adjusted for confounding factors (breastfeeding, delivery mode, antibiotic exposures, gestational age at birth, and twin status).

The infant gut microbiota taxa and alpha diversity measures at each sampling time that were most highly predictive of BMI z-score at age 12. Infant gut microbiota taxa and diversity measures (OTUs listed with OTU, phylum, and most specific level of known taxonomic classification) over the first 2 years of life that were selected as most predictive of childhood BMI z-score in random forests at six points (grouped vertically). We used linear regressions in order to assess the direction of association between these features at each sampling time with BMI z-score at age 12, controlling for confounding factors (breastfeeding, delivery mode, antibiotic exposures, gestational age at birth, and twin status). While a feature may have been selected as predictive of BMI only at a certain sampling time, we plotted the linear associations at all times (horizontal axis) in order to assess the temporal consistency of the association. Since the assumptions underlying random forests and linear regressions are very different, we would not expect important features to necessarily be significant in regressions. The colors represent the regression coefficients for each feature: red indicates a positive relationship between the feature and childhood BMI, e.g., higher abundance corresponds with higher BMI; green indicates negative relationships; gray indicates that the regression model failed to converge (NA). The column labeled “maternal taxa” shows whether these gut microbiota species or diversity measures were also associated with maternal overweight/obesity (Ow/Ob), excessive gestational weight gain (GWG), or both, in the maternal gut microbiota at the time of delivery; Fig. 5 provides more detail and Fig. S4 shows the association between these BMI-associated infant taxa and maternal characteristics.

Spaghetti plots of BMI z-scores over time by Ow/Ob status at age 12. The regression lines (denoted by thicker lines) show that BMI z-scores were fairly constant between ages 2 and 12 for children who were not Ow/Ob at age 12, but there was an increase in BMI z-scores during this time for children who became Ow/Ob. At age 2, there was no significant difference between BMI z-scores of children who later became Ow/Ob and those who did not.