Meta-analysis of effects of exclusive breastfeeding on infant gut microbiota across populations

Abstract

Previous studies on the differences in gut microbiota between exclusively breastfed (EBF) and non-EBF infants have provided highly variable results. Here we perform a meta-analysis of seven microbiome studies (1825 stool samples from 684 infants) to compare the gut microbiota of non-EBF and EBF infants across populations. In the first 6 months of life, gut bacterial diversity, microbiota age, relative abundances of Bacteroidetes and Firmicutes, and predicted microbial pathways related to carbohydrate metabolism are consistently higher in non-EBF than in EBF infants, whereas relative abundances of pathways related to lipid metabolism, vitamin metabolism, and detoxification are lower. Variation in predicted microbial pathways associated with non-EBF infants is larger among infants born by Caesarian section than among those vaginally delivered. Longer duration of exclusive breastfeeding is associated with reduced diarrhea-related gut microbiota dysbiosis. Furthermore, differences in gut microbiota between EBF and non-EBF infants persist after 6 months of age. Our findings elucidate some mechanisms of short and long-term benefits of exclusive breastfeeding across different populations.

Fig. 1

Effects of non-EBF vs. EBF on gut microbial diversity in infants ≤6 months of age. a Gut microbial alpha diversity (standardized Shannon index) by breastfeeding status by infant age at stool sample collection from each included studies. Fitted lines and 95% confidence intervals (95% CI) were from generalized additive mixed models (GAMM). b The difference in gut microbial alpha diversity (standardized Shannon index) between non-EBF and EBF infants ≤6 months of age from each study and the pooled effect across seven included studies (meta-analysis) with 95% CI. c The trend effect of gut microbial alpha diversity (standardized Shannon index) across EBF, non-EBF, and non-BF infants ≤6 months of age from each study and the pooled effect across five included studies (meta-analysis) with 95% CI. Data from Haiti and South Africa studies were not included as there was no non-BF group. In each study, to roughly test for trends across breastfeeding categories, breastfeeding was coded as a continuous variable (EBF = 1, non-EBF = 2, and non-BF = 3). d Pooled estimates and 95% CI for the difference in (standardized) gut microbial alpha diversity (four common alpha diversity indexes) between non-EBF and EBF infants ≤6 months of age. e Pooled estimates and 95% CI for the trend effect of (standardized) gut microbial alpha diversity (four common alpha diversity indexes) across EBF, non-EBF, and non-BF infants ≤6 months of age. Estimates for diversity difference or trend and corresponding standard errors from each study were from linear mixed-effect models (longitudinal data) or linear models (non-longitudinal data) and were adjusted for infant age at sample collection. Pooled estimates of standardized diversity difference or trend and their 95% CI were from random-effects meta-analysis models based on the adjusted estimates and corresponding standard errors of all included studies. Pooled estimates with false discovery rate (FDR)-adjusted pooled p-values < 0.1 are shown as triangles. EBF exclusive breastfeeding, non-EBF non-exclusive breastfeeding, non-BF non-breastfeeding, USA United States of America, CA California, FL Florida, MA Massachusetts, MO Missouri, NC North Carolina, DD diversity difference, SE standard error, PD_whole_tree phylogenetic diversity whole tree

Fig. 2

Effects of non-EBF vs. EBF on gut microbiota age in infants ≤6 months of age. a Gut (standardized) microbiota age of infants ≤6 months of age by breastfeeding status by age of infants at stool sample collection from each of seven included studies. Fitted lines and 95% confidence intervals (95% CI) were from generalized additive mixed models (GAMM). b The difference in gut (standardized) microbiota age between non-exclusively breastfed (non-EBF) and EBF infants ≤6 months of age from each study and the pooled effect across seven included studies (meta-analysis) with 95% CI. c The trend of gut (standardized) microbiota age across EBF, non-EBF, and non-BF infants ≤6 months of age from each study and the pooled effect across five included studies (meta-analysis) with 95% CI. The Haiti and South Africa studies were not included, as there was no non-BF group in these two studies. In each study, to test for trend across breastfeeding categories, breastfeeding was coded as a continuous variable in the model (EBF = 1, non-EBF = 2, and non-BF = 3). Estimates for (standardized) microbiota age difference or trend and corresponding standard error from each study were from linear mixed-effect models (for longitudinal data) or linear models (for non-longitudinal data) and were adjusted for age of infants at sample collection. EBF exclusive breastfeeding, non-EBF non-exclusive breastfeeding, non-BF no breastfeeding, USA United States of America, CA California, FL Florida, MA Massachusetts, MO Missouri, NC North Carolina, MD microbiota age difference, SE standard error

Fig. 3

Effects of non-EBF vs. EBF on gut bacterial taxa abundances in infants ≤6 months of age. a Gut bacterial phyla: heatmap of log(odds ratio) (log[OR]) of relative abundances of all gut bacterial phyla between non-EBF and EBF infants for each study and forest plot of pooled estimates across all seven studies with 95% confidence intervals (95% CI). b Gut bacterial families: heatmap of log(OR) of relative abundances of all gut bacterial families between non-EBF and EBF infants for each study and forest plot of pooled estimates across all seven studies with 95% CI. All log(OR) estimates of each bacterial taxa from each study were from generalized additive models for location scale and shape (GAMLSS) with zero-inflated beta family (BEZI) and were adjusted for age of infants at sample collection. Pooled log(OR) estimates and 95% CI (forest plot) were from random-effects meta-analysis models based on the adjusted log(OR) estimates and corresponding standard errors of all included studies. Pooled log(OR) estimates with pooled p-values < 0.05 are in red, and those with false discovery rate (FDR)-adjusted pooled p-values < 0.1 are shown as triangles. Missing (unavailable) values are in white. EBF exclusive breastfeeding, non-EBF non-exclusive breastfeeding, OR odds ratio, USA United States of America, CA California, FL Florida, MA Massachusetts, MO Missouri, NC North Carolina

Fig. 4

Effects of non-EBF vs. EBF on gut bacterial pathway abundances in infants ≤6 months of age. a Meta-analysis of all infants in all seven included studies: heatmap of log(odds ratio) (log[OR]) of relative abundances of gut microbial KEGG pathways at level 3 between non-EBF and EBF infants for each study and forest plot of pooled estimates of all seven studies with 95% confidence intervals (95% CI). b Meta-analysis of vaginally born infants in four studies: heatmap of log(OR) of relative abundances of gut microbial KEGG pathways at level 3 between non-EBF and EBF infants for each study and forest plot of pooled estimates of four studies with 95% CI. Only four studies with available birth mode information (Canada, Haiti, USA [CA–MA–MO] and USA [CA–FL]) are included. c Meta-analysis of C-section born infants in four studies: heatmap of log(OR) of relative abundances of gut microbial KEGG pathways at level 3 between non-EBF and EBF infants for each study and forest plot of pooled estimates of four studies with 95% CI. Only four studies with available birth mode information (Canada, Haiti, USA (CA–MA–MO) and USA (CA–FL)) are included. All log(OR) estimates of each pathway from each study were from generalized additive models for location scale and shape (GAMLSS) with zero-inflated beta family (BEZI) and were adjusted for age of infants at sample collection. Pooled log(OR) estimates and 95% CI (forest plot) were from random-effects meta-analysis models based on the adjusted log(OR) estimates and corresponding standard errors of all included studies. Pooled log(OR) estimates with pooled p-values < 0.05 are in red and those with false discovery rate (FDR)-adjusted pooled p-values < 0.1 are shown as triangles. Only pathways with FDR-adjusted pooled p-value < 0.1 are shown. EBF exclusive breastfeeding, non-EBF non-exclusive breastfeeding, KEGG Kyoto Encyclopedia of Genes and Genomes, OR odds ratio, USA United States of America, CA California, FL Florida, MA Massachusetts, MO Missouri, NC North Carolina

Fig. 5

The continued effects of EBF on the infant gut microbiota up to 2 years of age. Data from Bangladesh study only. a The impact of duration of EBF (shorter than 2 months vs. longer than 2 months from birth) on gut microbiota age. b The impact of duration of EBF on gut bacterial family composition. c The effects of diarrhea (vs. no diarrhea) around the time of stool sample collection on gut microbiota age in infants with duration of EBF shorter than 2 months vs. longer than 2 months from birth. d The effects of diarrhea (vs. no diarrhea) around the time of stool sample collection on gut microbial diversity (Shannon index) in infants with duration of EBF shorter than 2 months vs. longer than 2 months from birth. e The effects of diarrhea (vs. no diarrhea) around the time of stool sample collection on gut bacterial taxa composition at the family level in infants with duration of EBF shorter than 2 months vs. longer than 2 months from birth. f The effects of diarrhea (vs. no diarrhea) around the time of stool sample collection on gut bacterial taxa composition at the family level in infants receiving no breastfeeding at the time of diarrhea vs. infants receiving breastfeeding at the time of diarrhea. Number of infants n = 50 (duration of EBF ≤ 2 months n = 30, duration of EBF > 2 months n = 20). Number of samples 0–2 years of age n_s = 996 (duration of EBF ≤ 2 months n_s = 580, duration of EBF > 2 months n_s = 416). Number of samples 6 months to 2 years of age n_s = 674 (duration of EBF ≤ 2 months n_s = 378 [diarrhea n_s = 29, no diarrhea n_s = 349]; duration of EBF > 2 months n_s = 296 [diarrhea n_s = 19, no diarrhea n_s = 277]; with breastfeeding n_s = 616 [diarrhea n_s = 45, no diarrhea n_s = 571]; without breastfeeding n_s = 44 [diarrhea n_s = 2, no diarrhea n_s = 42]). Fitted lines and 95% confidence intervals (95% CI) were from generalized additive mixed models (GAMM). Gray dashed lines demarcate time periods tested. Black stars indicate statistical significance. EBF exclusive breastfeeding, BF breastfeeding