Infant Colic Represents Gut Inflammation and Dysbiosis

Abstract

Objective: To dissect potential confounding effects of breast milk and formula feeding on crying + fussing, fecal calprotectin, and gut microbiota in babies with colic. We hypothesized that infant colic is associated with gut inflammation linked to intestinal dysbiosis.

Study design: A nested case-control design of 3 of our studies was used to analyze clinical and laboratory data at presentation, comparing babies with colic with controls. All investigators other than the biostatistician were blinded during data analysis. Subjects were recruited based on their age and crying + fussy time. We screened 65 infants, 37 with colic, as defined by Barr diary (crying + fussing time >3 hours daily), who were compared with 28 noncolicky infants.

Results: Fecal calprotectin was elevated in babies with colic. For each mode of infant feeding (breast milk, formula, or breast + formula), infants' fecal calprotectin was higher in babies with colic. Infants with colic had similar levels of fecal alpha diversity (richness) when compared with controls, and alpha diversity was lower in breast-fed babies. Beta diversity at the phylum level revealed significant differences in microbial population. A phylum difference resulted from reduced Actinobacteria (95% of which are Bifidobacilli) in babies with colic. Species significantly associated with colic were Acinetobacter and Lactobacillus iners.

Conclusions: Colic is linked with gut inflammation (as determined by fecal calprotectin) and dysbiosis, independent of mode of feeding, with fewer Bifidobacilli.

Trial registration: Clinicaltrials.gov: NCT01279265 and NCT01849991.

Keywords: calprotectin; crying; functional bowel disorder; intestinal inflammation; microbiota; newborn.

Figure 1.

CONSORT flow diagram showing recruitment and data analysis.

Figure 2.

Relationship between fecal calprotectin, colic, and feeding modality in infants. A, Infants with colic (n = 29) have significantly higher levels of fecal calprotectin than normal controls (n = 25; P< .0001, Wilcoxon rank-sum test). B, Infants with colic had higher levels of calprotectin than controls when separated by feeding group (P= .019, Kruskal-Wallis with post hoc Dunn test and correction for multiple comparisons).

Figure 3.

Relationship between fecal calprotectin and crying + fussing time. Fecal calprotectin levels and crying + fussing time data were normalized by square root transformation and fit in linear models separately for subjects with and without colic. There was no significant relationship between calprotectin levels and crying + fussing time among the noncolic group (▴ symbol; solid regression line; coefficient, 0.66; 95% CI, −0.24 to 1.57; P= .144) and among children with colic (● symbol; dashed regression line; coefficient, −0.07; 95% Cl, −0.68 to 0.53; P = .809). Shaded area indicates 95% CI of the regression line.

Figure 4.

There was no relationship found between age and calprotectin level (P = .403; R² = −0.005).

Figure 5.

Fecal microbial community composition in infants with and without colic. A, Alpha diversity as determined by number of species, Shannon or Simpson diversity index. Colic samples (n = 37) are not significantly different from controls (n = 28). B, Alpha diversity differs significantly by feeding group (breastfed, n = 26; breast-formula, n = 11; formula n = 28). Number of species: breastfed vs formula, P= .022; breast-formula vs formula, P= .022; Shannon index: breastfed vs formula, P= .001; breastformula vs formula, P= .010; Simpson index: breastfed vs formula, P=.002, breast-formula vs formula, P= .012 (pairwise Wilcoxon rank sum tests with correction for multiple comparisons). C, Phylum/class level composition of colic (n = 37) vs control (n = 28) samples. Normal control samples have a significantly higher abundance of Actinobacteria (P= .032, Wilcoxon rank sum tests with correction for multiple comparisons). D, Microbial beta diversity composition of infants with and without colic is significantly different (P= .003, permutational multivariate analysis of variance).

Figure 6.

Differences at the species level between controls and babies with colic. Pairwise 2-sided Wilcoxon rank sum tests with subsequent correction for multiple comparisons showed 2 operational taxonomic units to be significantly different between the colic and control groups: Lactobacillus iners (P = .002) and Acinetobacter spp (P = .014).

Figure 7.

Random Forest plot of the most important variables, comparing infants with colic and their controls. Mean decrease in the Gini coefficient (>0.5) of attributes as assigned by the random forest model.