Table 3.

Multivariable regression models^a to estimate modification of IYCF on stunting at 18 mo by infant gut microbiome species in 53 infants in whom fecal specimens were available at the 6-month follow-up visit and stunting status was ascertained at the 18-month follow-up visit.

	n/Nb	PD (95% CI)	p-value	Adjusted p-valuec
Bifidobacterium longum
IYCF	17/53	−21.7% (−46.9%, 3.5%)	0.092	0.482
B.longum		−24.0% (−43.3%, −4.8%)	0.015	0.097
B.longum-by-IYCFd		49.1% (24.6%, 73.6%)	<0.001	0.002
Bifidobacterium pseudocatenulatum
IYCF	17/53	−21.9% (−49.5%, 6.0%)	0.12	0.503
B.pseudocatenulatum		37.5% (9.7%, 65.2%)	0.008	0.082
B.pseudocatenulatum-by-IYCFd,e		−47.8% (−75.7%, −20.0%)	0.001	0.005
Escherichia coli
IYCF	17/53	−1.9% (−29.1%, 25.3%)	0.889	0.966
E.coli		0.3% (−11.7%, 12.3%)	0.961	0.985
E.coli-by-IYCFd		−8.3% (−22.6%, 5.9%)	0.253	0.716
Dorea longicatena
IYCF	17/53	−4.1% (−31.0%, 22.8%)	0.766	0.966
D.longicatena		−0.1% (−10.6%, 10.4%)	0.985	0.985
D.longicatena-by-IYCFd		5.0% (−32.0%, 42.0%)	0.791	0.890
Dorea formicigenerans
IYCF	17/53	−17.5% (−35.9%, 0.9%)	0.062	0.434
D.formicigenerans		−17.1% (−33.0%, −1.2%)	0.035	0.174
D.formicigenerans-by-IYCFd		−33.6% (−79.1%, 11.8%)	0.147	0.528

PD, prevalence difference (in percentage points); 95% CI, 95% confidence interval; IYCF, infant and young child feeding.

^aCovariates include birthweight, infant sex, age at the 6 mo visit, infant LAZ at the 6 mo visit, infant diet diversity score at the 6mo visit, and mother-infant Lewis-null secretor phenotype discordance coded as both, none, infant only or mother only.

^bNumber stunted/Total N included in the model. Species relative abundances are centered and scaled to standard deviations and modelled as continuous variables.

^cAdjusted for multiple hypothesis testing by the Benjamini-Hochberg method.

^dPrevalence difference-in-differences per 1 SD increase in relative abundance.

^eWinsorized to the 90th percentile prior to centering and scaling.