Diet, adiposity, and the gut microbiota from infancy to adolescence: A systematic review

Abstract

Background:

Early life gut microbiota are affected by several factors that make identification of microbial-adiposity relationships challenging. This review evaluates studies that have investigated the gut microbiota composition associated with adiposity in infants, children, and adolescents and provides evidence-based nutrition recommendations that address microbiota-adiposity links.

Methods:

Electronic databases were systematically searched through January 2020. Eligible studies were published in English and analyzed gut microbiota and adiposity among individuals aged birth to 18 y. Abstracts and full-text articles were reviewed by three independent reviewers.

Results:

Of 45 full-text articles reviewed, 33 were included. No difference in abundance was found for Bacteroidetes (n=7/15 articles), Firmicutes (n=10/17) Actinobacteria (n=8/12), Proteobacteria (n=8/12), Tenericutes (n=4/5), and Verrucomicrobia (n=4/6) with adiposity. Lower abundance of Christensenellaceae (n=3/5) and Rikenellaceae (n=6/8) but higher abundance of F. prausnitzii (n=3/5) and Prevotella (n=5/7) were associated with adiposity.

Conclusions:

A lack of consensus exists for gut microbial composition associations with adiposity. A healthy gut microbiota is associated with a diet rich in fruits and vegetables with moderate consumption of animal fat and protein. Future research should use more robust sequencing technologies to identify all bacterial taxa associated with adiposity and evaluate how diet effects these adiposity-associated microbes.

INTRODUCTION

Pediatric obesity is a public health burden associated with metabolic syndrome,^1–3 type 2 diabetes,^4–6 and cardiovascular disease.¹ Data indicate that obesity during childhood and adolescence trends into adulthood,⁷ increasing the risk of obesity-related diseases across the lifespan. Obesity rates vary by age, affecting 13.9%, 18.4% and 20.6% of youth aged two to five, six to 11, and 12 to 19 years old, respectively.^8,9 Lifestyle modifications are the cornerstone approach to obesity prevention;^10–13 however, the gut microbiome (microorganisms that inhabit the human intestine) may also play a role. While each human has a unique microbiome composition that emerges as early as infancy,¹⁴ specific microbes have been shown to associate with specific health conditions.¹⁵ Therefore, it is imperative to understand how these novel differences in microbial community structure contribute to the development of adiposity across the lifespan to potentially target the modulation of the human gut microbiome in obesity prevention strategies.

Infancy is a period of rapid growth where microbial colonization of the gut begins through complex interactions between an individual’s genetics, environment, and behaviors.¹⁶ Microbiome disruption during infancy may lead to altered metabolic processes that continue throughout life.^17,18 The diet of an infant is an important modifiable factor that may result in microbial dysbiosis or aid in the development of a healthy gut microbiome. Breastfed infants exhibit a higher proportion of Bifidobacterium, bacterial species that ferment the oligosaccharides in human breast milk.¹⁹ Whereas, formula-fed infants have significantly higher proportions of bacteria from the phyla Firmicutes and Bacteroidetes, which are associated with overweight and obesity.¹⁴ Furthermore, infancy coincides with the transition from liquid to solid foods which results in a reduction of Bifidobacterium and an increase in Firmicutes.²⁰ The cessation of breastfeeding and the introduction of solid foods contribute to the functional maturation of the infant gut microbiome.²¹ If timed improperly, these important feeding events may promote microbial dysbiosis and a higher risk of obesity later in life.

The gut microbiome continues to develop during childhood and begins to stabilize around five years of age.^22,23 The dominant bacterial genera found in the healthy gut microbiome of children include Bacteroides, Prevotella and Bifidobacterium.²⁴ Childhood is a critical period of development when the gut microbiome begins to resemble that of an adult.²⁴ Dietary choices made during this time continue to shape the evolving microbiome.²⁵ Western-style diets have been shown to decrease Bifidobacterium and increase Eubacterium in the gut microbiome,²⁶ associating with a greater prevalence of overweight or obesity.^27,28 Differences in the gut microbiome structure have been observed where children with obesity exhibit a higher ratio of Firmicutes-to-Bacteroidetes (F:B ratio), higher amounts of Lactobacillus spp., and lower abundance of Bacteroides vulgatus as compared to lean children.²⁹ This provides further evidence that microbial dysbiosis may contribute to the occurrence of childhood obesity.

Adolescence is characterized by a secondary period of rapid growth but also marks a period of maturation and transition to greater independence. Previous studies have found that the adolescent gut microbiome has significantly higher abundance of Bifidobacterium and Clostridium when compared to the healthy adult gut microbiome.³⁰ Typically, Bifidobacterium is in higher abundance during infancy and Clostridium is more abundant during adulthood;³⁰ however, the simultaneous presence of these bacteria may either be reflective of the important contribution of this developmental stage to the structure of the gut microbiome community or the functional need for microbial metabolites during continued growth and development. Among adolescents with obesity, greater Actinomyces abundance has been observed.³¹ Shifts in bacterial composition during adolescence to a more stable, adult composition is important for disease prevention as microbial changes could have a lasting effect on health into adulthood.

With such a wealth of microbes in the gut microbiome and the dynamic nature of growth and development, it is challenging to identify microbes that co-exist and potentially contribute to obesity in early life. The purpose of this systematic review was to evaluate studies that have investigated gut bacteria that are associated with adiposity in infants, children, and adolescents. Additionally, this review provides evidence-based dietary approaches that have been shown to impact the obesity-associated gut microbes. Additionally, we provide best practices for future research evaluating microbial relationships with adiposity during these formative years.

METHODS

The protocol was registered through PROSPERO (Registration #: CRD42018108054) and data extraction followed the 2009 Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines.³²

Search Strategy

Studies were identified using a standardized search strategy to evaluate articles that assessed differences in the gut microbiome among lean infants, children, and adolescents as compared to those with overweight or obesity. Articles that assessed dietary components associated with microbial composition were identified to provide nutritional recommendations that have shown beneficial effects on microbial composition. The electronic databases (PubMed, CINAHL, and Web of Science) were searched from inception to January 2020 using a combination of the following Medical Subject Headings terms: pediatric obesity, obesity, adiposity, gastrointestinal microbiome, microbiota, child, child preschool, and/or adolescent.

Inclusion and Exclusion Criteria

Inclusion criteria included: (1) randomized or non-randomized trials, crossover, cross-sectional, parallel design, and prospective observational studies; (2) collection and analyses of gut microbiome through fecal samples; (3) at least one adiposity measure; (4) participants aged 0–18 years old; and (5) published in English. Exclusion criteria included all genetic diseases and conditions associated with overweight or obesity.

Data Extraction

A standardized data form (Microsoft Excel, Microsoft, 2010) was used for extraction and tabulation. Two independent reviewers (K.B.V and C.P.O) evaluated the titles and abstracts to identify eligible articles. Full texts of selected articles were reviewed for inclusion and exclusion criteria with reference lists evaluated for additional eligible articles. Disagreements were resolved with a third reviewer (C.M.W) prior to final inclusion decision.

Assessment of Quality

The Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies developed by the National Heart, Lung, and Blood Institute³³ was used to assess article quality. This 14-item tool evaluates the internal validity and potential flaws in methodology and implementation of cohort and cross-sectional studies. A quality rating of “good” (yes to >7 items), “fair” (yes to 4–6 items), or “poor” (yes to <3 items) was assigned to each article. For items where “no” was selected, reviewers identified the category of potential risk. One independent reviewer (K.B.V) provided responses for all 14 items and an overall quality rating. A secondary reviewer (S.N.T.) reviewed responses and overall quality ratings and any discrepancies were resolved by a third reviewer (C.M.W.).

Assessment of Potential Bias

Identification of bias was based on the Research Triangle Institute (RTI) 29-item bank created by RTI International for evaluation of risk of bias and precision in observational studies.³⁴ One independent reviewer (K.B.V) identified potential sources of bias based on the answers to item bank questions that were applicable to each study design. A secondary reviewer (C.P.O) assessed the responses provided by the first reviewer and any discrepancies were resolved by a third reviewer (C.M.W.).

RESULTS

Study Selection

During the search, 2096 studies were identified, of which 1091 were duplicate articles across search strings and electronic databases. Of the 1696 additional titles screened from reference lists, none were eligible for inclusion. Figure 1 shows the total number of screened, excluded, and included full-text articles.

Figure 1.

PRISMA Flow Diagram

Study Characteristics

Supplemental Table 1 summarizes the main characteristics and conclusions of the included articles. Among the included studies, 82% (n=27) were cross-sectional studies^{29,31,35–59} with a mean sample size of 316. The location of the research also varied with seven from Mexico,^{35,36,38–40,55,59} six from the USA,^{42,45,50,51,60,61} four from China,^46,52,53,62 three from Italy,^31,49,63 and one each from Belgium,²⁹ Brazil,⁵⁶ Canada,⁶⁴ Egypt,⁴⁸ Germany,⁴⁴ India,⁴³ Iran,³⁷ Korea,⁴⁷ New Zealand,⁵⁸ Malawi,⁶⁵ Sweden,⁵⁷ Switzerland,⁴¹ and Trinidad.⁵⁴ Forty-two percent (n=14)^{35,36,39,42,45,46,53,54,58–61,64,65} of the studies used 16S rRNA sequencing alone for microbial analysis with five^{40,43,52,55,62} using quantitative PCR alone and nine^{37,38,47–51,56,57} using 16S rRNA sequencing with quantitative PCR. The remaining five^{29,31,41,44,63} studies used a combination of other methods to asses microbial differences. Various methods of categorizing participants by weight status were used with the most common method (n=12) being body mass index (BMI) percentiles for determination of categorical cut-offs.^{31,36,38,39,42,43,45,51,52,54,59,61} There was one study that did not provide a definition used for the classification of children with obesity vs. healthy children.⁴⁶ Fifty-three percent (n=3622/6816) of the included study cohorts were male. Four studies^38,46,48,65 did not provide the sex breakdown for their cohort and were not included in the overall percent. Only two studies^36,60 reported the gestational age at delivery for their cohorts with both being approximately 39 weeks.

Summary of Existing Literature

The main findings regarding microbial relationships with adiposity among the included studies are summarized below and in Table 1. This table includes the 26 microbes that were reported in four or more of the included articles, the association with adiposity, age group of each study cohort, and the references for the articles. Supplemental Table 2 includes all 70 microbes that were reported among the included studies. Figure 2A summarizes the 21 microbes that were reported in five or more of the included studies and the percent agreement with the indicated adiposity relationship. The proceeding summary is grouped by developmental period (i.e. infants/toddlers, school-aged children, and adolescents). There were several studies that had age ranges that spanned several stages of development.^{29,35–42,46,48,49,52–56,59,62,63} For those studies, the stage of development was determined using the mean age of the study population.

Table 1.

Microbial Relationships with Adiposity Reported among Four or More of the Included Studies

	Microbe	Association with Adiposity		Age Group	Reference
		%	Relationship
Phylum	Actinobacteria	16.7%	Increase/Positive correlation	Adolescents	Del Chierico et al, Frontiers in Microbiology. 2018;9:1–12 (31). Ferrer et al, Environmental Microbiology. 2013;15:211–26 (44).
		16.7%	Decrease/Negative correlation	School-aged children	Chen et al, PeerJ. 2020;8:e8317 (53)
				Adolescents	Yuan et al. J of Gastroenterology and Hepatology. 2014;29(6):1292–1298 (51).
		66.7%	No difference/No association	Infant/Toddler	Forbes et al, JAMA Pediatrics. 2018;172:e181161 (64). Karvonen et al. International J of Obesity. 2019;43:713–723 (45).
				School-aged children	Da Silva et al, Childhood Obesity. 2020;16(3):204–210 (54). Méndez-Salazar et al, Frontiers in Microbiology. 2018;9(Oct):1–11 (35). Murugesan et al. Euro J of Clin Micro & Infect Diseases. 2015;34(7):1337–1346 (38). Nirmalkar et al, Nutrients. 2018;10(2009):1–23 (39).
				Adolescents	Nirmalkar et al, Nutrients. 2018;10(2009):1–23 (39). Riva et al. Environmental Microbiology. 2017;19(1):95–105 (49).
	Bacteroidetes	13.3%	Increase/Positive correlation	Adolescents	Del Chierico et al, Frontiers in Microbiology. 2018;9:1–12 (31). Yuan et al. J of Gastroenterology and Hepatology. 2014;29(6):1292–1298 (51).
		40.0%	Decrease/Negative correlation	School-aged children	Chen et al, PeerJ. 2020;8:e8317 (53) Méndez-Salazar et al, Frontiers in Microbiology. 2018;9(Oct):1–11 (35). Orbe-Orihuela et al. Salud Publica de Mexico. 2018;60(1):5–11 (40). Xu et al. BMC Microbiology. 2012;12(1):283–289 (62).
				Adolescents	Ferrer et al, Environmental Microbiology. 2013;15:211–26 (44). Riva et al. Environmental Microbiology. 2017;19(1):95–105 (49).
		46.7%	No difference/No association	Infant/Toddler	Forbes et al, JAMA Pediatrics. 2018;172:e181161 (64). Karvonen et al. International J of Obesity. 2019;43:713–723 (45).
				School-aged children	Da Silva et al, Childhood Obesity. 2020;16(3):204–210 (54). Mousavi et al. Iranian Red Crescent Med Journal. 2018;20(4):1–6 (37). Murugesan et al. Euro J of Clin Micro & Infect Diseases. 2015;34(7):1337–1346 (38). Nirmalkar et al, Nutrients. 2018;10(2009):1–23 (39).
				Adolescents	Nirmalkar et al, Nutrients. 2018;10(2009):1–23 (39).
	Firmicutes	35.3%	Increase/Positive correlation	School-aged children	Da Silva et al, Childhood Obesity. 2020;16(3):204–210 (54). Mousavi et al. Iranian Red Crescent Med Journal. 2018;20(4):1–6 (37). Orbe-Orihuela et al. Salud Publica de Mexico. 2018;60(1):5–11 (40).
				Adolescents	Del Chierico et al, Frontiers in Microbiology. 2018;9:1–12 (31). Ferrer et al, Environmental Microbiology. 2013;15:211–26 (44). Riva et al. Environmental Microbiology. 2017;19(1):95–105 (49).
		5.9%	Decrease/Negative correlation	Adolescents	Yuan et al. J of Gastroenterology and Hepatology. 2014;29(6):1292–1298 (51).
		58.8%	No difference/No association	Infant/Toddler	Forbes et al, JAMA Pediatrics. 2018;172:e181161 (64). Karvonen et al. International J of Obesity. 2019;43:713–723 (45).
				School-aged children	Chen et al, PeerJ. 2020;8:e8317 (53) López-Contreras et al. Pediatric Obesity. 2017;13(6):381–388 (59). Méndez-Salazar et al, Frontiers in Microbiology. 2018;9(Oct):1–11 (35). Nirmalkar et al, Nutrients. 2018;10(2009):1–23 (39). Payne et al. Nutrition & Diabetes. 2011;1(7):e12 (41). Xu et al. BMC Microbiology. 2012;12(1):283–289 (62). Murugesan et al. Euro J of Clin Micro & Infect Diseases. 2015;34(7):1337–1346 (38).
				Adolescents	Nirmalkar et al, Nutrients. 2018;10(2009):1–23 (39).
	Firmicutes:Bacteroidetes	41.7%	Increase/Positive correlation	School-aged children	Mousavi et al. Iranian Red Crescent Med Journal. 2018;20(4):1–6 (37). Xu et al. BMC Microbiology. 2012;12(1):283–289 (62).
				Adolescents	Bervoets et al, Gut Pathogens. 2013;5:1–10 (29) Hou et al, BioMed Research International. 2017;7585989:1–8 (46). Riva et al. Environmental Microbiology. 2017;19(1):95–105 (49).
		58.3%	No difference/No association	Infant/Toddler	Forbes et al, JAMA Pediatrics. 2018;172:e181161 (64). Craig SJC et al. Scientific Reports. 2018;8:14030 (60).
				School-aged children	Hollister et al, Childhood Obesity. 2018;14(2):122–130 (61). López-Contreras et al. Pediatric Obesity. 2017;13(6):381–388 (59). Chen et al, PeerJ. 2020;8:e8317 (53)
				Adolescents	Hu et al, PLoS ONE. 2015;10(7):1–14 (47). Whisner et al, BMC Microbiology. 2018;18(1)1:11 (50).
	Proteobacteria	25.0%	Increase/Positive correlation	School-aged children	Chen et al, PeerJ. 2020;8:e8317 (53) Méndez-Salazar et al, Frontiers in Microbiology. 2018;9(Oct):1–11 (35).
				Adolescents	Yuan et al. J of Gastroenterology and Hepatology. 2014;29(6):1292–1298 (51).
		8.3%	Decrease/Negative correlation	Adolescents	Ferrer et al, Environmental Microbiology. 2013;15:211–26 (44).
		66.7%	No difference/No association	Infant/Toddler	Forbes et al, JAMA Pediatrics. 2018;172:e181161 (64). Karvonen et al. International J of Obesity. 2019;43:713–723 (45).
				School-aged children	Da Silva et al, Childhood Obesity. 2020;16(3):204–210 (54). Nirmalkar et al, Nutrients. 2018;10(2009):1–23 (39). Murugesan et al. Euro J of Clin Micro & Infect Diseases. 2015;34(7):1337–1346 (38).
				Adolescents	Nirmalkar et al, Nutrients. 2018;10(2009):1–23 (39). Del Chierico et al, Frontiers in Microbiology. 2018;9:1–12 (31). Riva et al. Environmental Microbiology. 2017;19(1):95–105 (49).
	Tenericutes	20.0%	Decrease/Negative correlation	School-aged children	Chen et al, PeerJ. 2020;8:e8317 (53)
		80.0%	No difference/No association	School-aged children	Méndez-Salazar et al, Frontiers in Microbiology. 2018;9(Oct):1–11 (35). Nirmalkar et al, Nutrients. 2018;10(2009):1–23 (39).
				Adolescents	Nirmalkar et al, Nutrients. 2018;10(2009):1–23 (39). Yuan et al. J of Gastroenterology and Hepatology. 2014;29(6):1292–1298 (51).
	Verrucomicrobia	33.3%	Decrease/Negative correlation	Infant/Toddler	Karvonen et al. International J of Obesity. 2019;43:713–723 (45).
				Adolescents	Hou et al, BioMed Research International. 2017;7585989:1–8 (46).
		66.7%	No difference/No association	Infant/Toddler	Forbes et al, JAMA Pediatrics. 2018;172:e181161 (64).
				Adolescents	Del Chierico et al, Frontiers in Microbiology. 2018;9:1–12 (31). Riva et al. Environmental Microbiology. 2017;19(1):95–105 (49). Yuan et al. J of Gastroenterology and Hepatology. 2014;29(6):1292–1298 (51).
Family	Bacteroidaceae	25.0%	Increase/Positive correlation	Adolescents	Ferrer et al, Environmental Microbiology. 2013;15:211–26 (44).
		50.0%	Decrease/Negative correlation	Adolescents	Hu et al, PLoS ONE. 2015;10(7):1–14 (47).
				Adolescents	Riva et al. Environmental Microbiology. 2017;19(1):95–105 (49).
		25.0%	No difference/No association	Infant/Toddler	Forbes et al, JAMA Pediatrics. 2018;172:e181161 (64).
	Christensenellaceae	20.0%	Increase/Positive correlation	School-aged children	Leong et al. Am J of Clinical Nutrition. 2020’111(1):70–78 (58).
		60.0%	Decrease/Negative correlation	School-aged children	Moran-Ramos et al, Gut Microbes. 2020; Jan 23:1–18 (36). López-Contreras et al. Pediatric Obesity. 2017;13(6):381–388 (59).
				Adolescents	Ferrer et al, Environmental Microbiology. 2013;15:211–26 (44).
		20.0%	No difference/No association	Adolescents	Riva et al. Environmental Microbiology. 2017;19(1):95–105 (49).
	Coriobacteriaceae	75.0%	Increase/Positive correlation	Infant/Toddler	Forbes et al, JAMA Pediatrics. 2018;172:e181161 (64).
				School-aged children	Nirmalkar et al, Nutrients. 2018;10(2009):1–23 (39).
				Adolescents	Nirmalkar et al, Nutrients. 2018;10(2009):1–23 (39).
		25.0%	No difference/No association	School-aged children	Da Silva et al, Childhood Obesity. 2020;16(3):204–210 (54).
	Enterobacteriaceae	28.6%	Increase/Positive correlation	Infant/Toddler	Kamng’ona et al. Scientific Reports;2019;9:12893 (65).
				School-aged children	Karlsson et al, Obesity. 2012;20(11):2257–61 (57).
		14.3%	Decrease/Negative correlation	School-aged children	López-Contreras et al. Pediatric Obesity. 2017;13(6):381–388 (59).
		57.1%	No difference/No association	Infant/Toddler	Forbes et al, JAMA Pediatrics. 2018;172:e181161 (64).
				School-aged children	Borgo F et al, Childhood Obesity. 2017;13(1):78–84 (63). Payne et al. Nutrition & Diabetes. 2011;1(7):e12 (41).
				Adolescents	Hu et al, PLoS ONE. 2015;10(7):1–14 (47).
	Lachnospiraceae	37.5%	Increase/Positive correlation	Infant/Toddler	Forbes et al, JAMA Pediatrics. 2018;172:e181161 (64).
				School-aged children	Murugesan et al. Euro J of Clin Micro & Infect Diseases. 2015;34(7):1337–1346 (38).
				Adolescents	Del Chierico et al, Frontiers in Microbiology. 2018;9:1–12 (31).
		62.5%	No difference/No association	School-aged children	Da Silva et al, Childhood Obesity. 2020;16(3):204–210 (54). Moran-Ramos et al, Gut Microbes. 2020; Jan 23:1–18 (36).
				Adolescents	Ferrer et al, Environmental Microbiology. 2013;15:211–26 (44). Hu et al, PLoS ONE. 2015;10(7):1–14 (47). Riva et al. Environmental Microbiology. 2017;19(1):95–105 (49).
	Prevotellaceae	25.0%	Increase/Positive correlation	Adolescents	Hu et al, PLoS ONE. 2015;10(7):1–14 (47).
		75.0%	No difference/No association	Infant/Toddler	Forbes et al, JAMA Pediatrics. 2018;172:e181161 (64).
				Adolescents	Ferrer et al, Environmental Microbiology. 2013;15:211–26 (44). Riva et al. Environmental Microbiology. 2017;19(1):95–105 (49).
	Rikenellaceae	77.8%	Decrease/Negative correlation	School-aged children	Moran-Ramos et al, Gut Microbes. 2020; Jan 23:1–18 (36). Chen et al, PeerJ. 2020;8:e8317 (53) López-Contreras et al. Pediatric Obesity. 2017;13(6):381–388 (59).
				Adolescents	Ferrer et al, Environmental Microbiology. 2013;15:211–26 (44). Hu et al, PLoS ONE. 2015;10(7):1–14 (47). Bai et al. Pediatric Obesity. 2019;14(4):e12480 (42).
		22.2%	No difference/No association	Infant/Toddler	Forbes et al, JAMA Pediatrics. 2018;172:e181161 (64).
				Adolescents	Riva et al. Environmental Microbiology. 2017;19(1):95–105 (49).
	Ruminococcaceae	50.0%	Increase/Positive correlation	Infant/Toddler	Forbes et al, JAMA Pediatrics. 2018;172:e181161 (64).
				School-aged children	Leong et al. Am J of Clinical Nutrition. 2020’111(1):70–78 (58).
				Adolescents	Riva et al. Environmental Microbiology. 2017;19(1):95–105 (49).
		33.3%	Decrease/Negative correlation	Adolescents	Ferrer et al, Environmental Microbiology. 2013;15:211–26 (44). Hu et al, PLoS ONE. 2015;10(7):1–14 (47).
		16.7%	No difference/No association	School-aged children	Da Silva et al, Childhood Obesity. 2020;16(3):204–210 (54).
	Veillonellaceae	50.0%	Increase/Positive correlation	School-aged children	Moran-Ramos et al, Gut Microbes. 2020; Jan 23:1–18 (36).
				Adolescents	Hu et al, PLoS ONE. 2015;10(7):1–14 (47).
		50.0%	No difference/No association	Infant/Toddler	Forbes et al, JAMA Pediatrics. 2018;172:e181161 (64).
				Adolescents	Riva et al. Environmental Microbiology. 2017;19(1):95–105 (49).
Genus	Bacteroides	57.1%	Decrease/Negative correlation	School-aged children	Leong et al. Am J of Clinical Nutrition. 2020’111(1):70–78 (58).
				Adolescents	Hou et al, BioMed Research International. 2017;7585989:1–8 (46). Hu et al, PLoS ONE. 2015;10(7):1–14 (47). Riva et al. Environmental Microbiology. 2017;19(1):95–105 (49).
		42.9%	No difference/No association	Infant/Toddler	Karvonen et al. International J of Obesity. 2019;43:713–723 (45).
				School-aged children	Payne et al. Nutrition & Diabetes. 2011;1(7):e12 (41). Moran-Ramos et al, Gut Microbes. 2020; Jan 23:1–18 (36).
	Bifidobacterium	37.5%	Increase/Positive correlation	School-aged children	Leong et al. Am J of Clinical Nutrition. 2020’111(1):70–78 (58).
				Adolescents	Nirmalkar et al, Nutrients. 2018;10(2009):1–23 (39). Bai et al. Pediatric Obesity. 2019;14(4):e12480 (42).
		12.5%	Decrease/Negative correlation	School-aged children	Da Silva et al, Childhood Obesity. 2020;16(3):204–210 (54).
		50.0%	No difference/No association	Infant/Toddler	Karvonen et al. International J of Obesity. 2019;43:713–723 (45).
				School-aged children	Karlsson et al, Obesity. 2012;20(11):2257–61 (57). Payne et al. Nutrition & Diabetes. 2011;1(7):e12 (41).
				Adolescents	Riva et al. Environmental Microbiology. 2017;19(1):95–105 (49).
	Blautia	40.0%	Increase/Positive correlation	Adolescents	Nirmalkar et al, Nutrients. 2018;10(2009):1–23 (39). Hou et al, BioMed Research International. 2017;7585989:1–8 (46).
		60.0%	No difference/No association	Infant/Toddler	Karvonen et al. International J of Obesity. 2019;43:713–723 (45).
				School-aged children	Da Silva et al, Childhood Obesity. 2020;16(3):204–210 (54).
				Adolescents	Riva et al. Environmental Microbiology. 2017;19(1):95–105 (49).
	Dialister	37.5%	Increase/Positive correlation	School-aged children	Moran-Ramos et al, Gut Microbes. 2020; Jan 23:1–18 (36). López-Contreras et al. Pediatric Obesity. 2017;13(6):381–388 (59).
				Adolescents	Bai et al. Pediatric Obesity. 2019;14(4):e12480 (42).
		25.0%	Decrease/Negative correlation	School-aged children	Chen et al, PeerJ. 2020;8:e8317 (53)
				Adolescents	Bai et al. Pediatric Obesity. 2019;14(4):e12480 (42).
		37.5%	No difference/No association	Adolescents	Hu et al, PLoS ONE. 2015;10(7):1–14 (47). Riva et al. Environmental Microbiology. 2017;19(1):95–105 (49).
	Faecalibacterium	25.0%	Increase/Positive correlation	School-aged children	Chen et al, PeerJ. 2020;8:e8317 (53)
				Adolescents	Bai et al. Pediatric Obesity. 2019;14(4):e12480 (42).
		25.0%	Decrease/Negative correlation	Adolescents	Hu et al, PLoS ONE. 2015;10(7):1–14 (47). Bai et al. Pediatric Obesity. 2019;14(4):e12480 (42).
		50.0%	No difference/No association	Infant/Toddler	Karvonen et al. International J of Obesity. 2019;43:713–723 (45).
				School-aged children	Da Silva et al, Childhood Obesity. 2020;16(3):204–210 (54). Leong et al. Am J of Clinical Nutrition. 2020’111(1):70–78 (58).
				Adolescents	Riva et al. Environmental Microbiology. 2017;19(1):95–105 (49).
	Lactobacillus	60.0%	Increase/Positive correlation	School-aged children	Da Silva et al, Childhood Obesity. 2020;16(3):204–210 (54). Nirmalkar et al, Nutrients. 2018;10(2009):1–23 (39).
				Adolescents	Bai et al. Pediatric Obesity. 2019;14(4):e12480 (42).
		40.0%	No difference/No association	School-aged children	Karlsson et al, Obesity. 2012;20(11):2257–61 (57). Payne et al. Nutrition & Diabetes. 2011;1(7):e12 (41).
	Prevotella	71.4%	Increase/Positive correlation	Infant/Toddler	Kamng’ona et al. Scientific Reports;2019;9:12893 (65).
				School-aged children	Moran-Ramos et al, Gut Microbes. 2020; Jan 23:1–18 (36). López-Contreras et al. Pediatric Obesity. 2017;13(6):381–388 (59).
				Adolescents	Nirmalkar et al, Nutrients. 2018;10(2009):1–23 (39). Hu et al, PLoS ONE. 2015;10(7):1–14 (47).
		14.3%	Decrease/Negative correlation	Adolescents	Bai et al. Pediatric Obesity. 2019;14(4):e12480 (42).
		14.3%	No difference/No association	Infant/Toddler	Karvonen et al. International J of Obesity. 2019;43:713–723 (45).
	Sutterella	50.0%	Increase/Positive correlation	School-aged children	Moran-Ramos et al, Gut Microbes. 2020; Jan 23:1–18 (36). Chen et al, PeerJ. 2020;8:e8317 (53)
				Adolescents	Hou et al, BioMed Research International. 2017;7585989:1–8 (46).
		33.3%	Decrease/Negative correlation	School-aged children	López-Contreras et al. Pediatric Obesity. 2017;13(6):381–388 (59).
				Adolescents	Bai et al. Pediatric Obesity. 2019;14(4):e12480 (42).
		16.7%	No difference/No association	Adolescents	Hu et al, PLoS ONE. 2015;10(7):1–14 (47).
Species	B. ovatus	25.0%	Increase/Positive correlation	School-aged children	López-Contreras et al. Pediatric Obesity. 2017;13(6):381–388 (59).
		25.0%	Decrease/Negative correlation	Adolescents	Mansour et al. Research J of Pharm, Biol, and Chem Sci. 2016;7(5):436–445 (48).
		50.0%	No difference/No association	School-aged children	Ignacio et al, Clin Microbial Infect. 2016;22:258.e1–258.e8 (56).
				Adolescents	Bervoets et al, Gut Pathogens. 2013;5:1–10 (29)
	F. prausnitzii	60.0%	Increase/Positive correlation	Adolescents	Del Chierico et al, Frontiers in Microbiology. 2018;9:1–12 (31). Balamurugan et al, British J of Nutrition. 2010;103:335–338 (43)
				Infant/Toddler	Kamng’ona et al. Scientific Reports;2019;9:12893 (65).
		20.0%	Decrease/Negative correlation	School-aged children	Borgo F et al, Childhood Obesity. 2017;13(1):78–84 (63).
		20.0%	No difference/No association	School-aged children	Payne et al. Nutrition & Diabetes. 2011;1(7):e12 (41).

Microbes are only included in the above table if a relationship was reported with adiposity in two or more of the included articles. The percent column is the percentage of articles in agreement for each indicated association with adiposity by age group. The denominator used to calculate each percentage is the number of articles that reported a relationship with adiposity for that specific microbe. Note: if a study included data for two age groups, then that study was included twice in the denominator for a specific microbe (e.g. the denominator for Actinobacteria is 12)

Figure 2.

Percent agreement of microbial associations with adiposity outcomes among the articles included in the current review by specific microbe for (A) entire cohort regardless of age group, (B) infants and toddlers, (C) school-aged children, and (D) adolescents. Black indicates no association/difference, dark gray indicates a significantly lower abundance or negative association, and light gray indicates significantly higher abundance or positive association. Microbial names are indented to show related taxa at lower taxonomic levels. Bolded microbial names are phyla-level bacteria. Percentage next to microbial name indicates the percent of the included articles that reported this specific microbe. Panel A: Only microbes that were reported in five or more of the included articles. Panels B-D: Only microbes reported in two or more of the included articles for the indicated developmental period.

Infancy/Toddlerhood (Birth to < 4 years).

Of the included studies, only 12% (n=4) evaluated gut microbial composition during infancy and toddlerhood in association with adiposity. Three of these studies used longitudinal, observational^60,64,65 study designs with one study using a cross-sectional⁴⁵ design. The first three studies^60,64,65 were of good quality with the last study being of fair study quality;⁴⁵ however, all studies adjusted for multiple comparisons and controlled for several covariates. Figure 2B provides a summary of the key microbial relationships with adiposity for the four included articles that had infant/toddler-aged populations.

The first study found that the child’s gut microbiota was not associated with weight gain during the first two years of life.⁶⁰ However, there were distinct clusters of both Bacteroidetes and Firmicutes that were positively associated with growth, whereas Proteobacteria was negatively associated with growth.⁶⁰ This study also found a discriminant group within the Actinobacteria phyla that was more abundant in children without rapid weight gain.⁶⁰ Additional findings from this study suggested that microbial obesity signatures were more evident in the oral microbiome (vs. gut microbiome) during infancy and toddlerhood and that obesity-associated gut microbiota shifts may become more pronounced in later stages of life.⁶⁰

The second study found that the composition of the gut microbiota partially explained a strong, inverse relationship between breastfeeding and risk of overweight.⁶⁴ It was reported that infants partially breastfed or exclusively formula-fed at 3 months of age had a 63% and 102% greater risk, respectively, of being overweight at 12 months of age when compared to exclusively breastfed infants.⁶⁴ This study found significantly higher relative abundance of Coriobacteriaceae, Erysipalotrichaceae, Lachnospiraceae, and Ruminococcaceae at 3–4 months of age among infants who became overweight by 12 months.⁶⁴ Subsequently, the study determined that increasing exclusivity of breastfeeding (formula-fed, partial, exclusive) was associated with decreased relative abundance of Lachnospiraceae and Ruminococcaceae.⁶⁴ This study provides support that breastfeeding may be protective of overweight via the early gut microbiome.⁶⁴

The third study found that gut microbiome maturity and diversity at 6 months was associated with weight but not linear growth between 6–12 months of age; however, this relationship did not persist at 12–18 months of age.⁶⁵ Overall, this study found that Proteobacteria and Bacteroidetes were positively associated whereas Actinobacteria were negatively associated with weight gain.⁶⁵ It was reported that weight-for-age z-score was positively associated with Prevotella, Campylobacter, Enterobacter ludwigii, Klebsiella, and Salmonella enterica and negatively associated with Actinomyces, Atopobium parvulum, Streptococcus, and C. difficile.⁶⁵ This provides evidence that specific microbes may play a key role in weight gain during infancy.

Lastly, the cross-sectional study found significantly lower microbial abundance of Verrucomicrobia, Ruminococcus, Akkermansia, Parabacteroidetes but significantly higher abundance of Dorea among three-year old toddlers with greater adiposity.⁴⁵ This is similar to studies in older children and adolescents^44,46,47 that are described in later paragraphs and suggests that early life changes in the gut microbiome may predispose an individual to obesity later in life.

School-aged Children (>4 but < 11 years).

Of the included studies, 52% (n=17) evaluated the microbial composition associated with adiposity among school-aged children. The majority (82%, n=14) were cross-sectional with the remaining studies using case-control (n=2) and longitudinal (n=1) study designs. Ninety-three percent (n=13)^{35–41,53,54,56–59} of the cross-sectional studies were of fair study quality with the remaining cross-sectional study being of good study quality.⁵⁵ Six^{36,40,55,56,58,59} of these studies controlled for potential confounding variables and seven^{35,36,39,41,55,56,59} adjusted for multiple comparisons. Both of the case-control studies were of fair study quality and controlled for multiple comparisons.^62,63 Figure 2C provides a summary of the key microbial relationships with adiposity for the 17 included articles among school-aged children.

There were major differences in the results among the cross-sectional studies. Four studies reported that the relative abundance of Bacteroidetes was significantly lower in children with overweight/obesity,^35,40,53,62 whereas four studies found no difference in Bacteroidetes abundance by weight status.^37–39,54 Similarly, there were mixed results regarding the abundance of Firmicutes with six studies reporting no difference^{35,38,39,41,53,59,62} and three studies reporting significantly higher abundance^37,40,54 in children with overweight/obesity. These findings led to variability regarding F:B ratio, with three studies reporting no difference^53,59,61 and two studies finding a significantly higher ratio^37,62 among children with obesity compared to normal-weight controls. Furthermore, there were four studies that found no differences^35,38,39,54 and only one study⁵³ that found significantly lower levels of Actinobacteria abundance among children with greater adiposity. Lastly, three studies found no difference^38,39,54 in Proteobacteria whereas two studies^35,53 reported significantly greater abundance among children with overweight/obesity compared to normal-weight children.

At the class level, two studies reported significantly higher abundance of Betaproteobacteria with greater adiposity.^36,53 At the family level, three studies reported conflicting results for Enterobacteriaceae^41,57,59 and Christensenellaceae.^36,58,59 Three studies found significantly lower levels of Oscillospira.^36,38,53 Further evaluation at the genus level revealed a lack of consensus for Bifidobacterium,^41,54,56,57 Lactobacillus,^{39,41,54,56,57} and Faecalibacterium,^38,53,54,58 with two studies reporting that children with obesity had elevated Clostridium.^39,56 At the species level, Bacteroides fragilis^56,57 and ovatus^56,59 abundance were inconclusive but two studies reported a significantly lower abundance of Bacteroides plebeius.^59,61

One of the case-control studies found a lower Bacteroidetes abundance but significantly higher F:B ratio⁶² in children with obesity, which was similar to two of the cross-sectional studies.^37,40 The other case-control study evaluated microbial abundance at the species level finding significantly lower levels of Akkermansia muciniphila and Faecalibacterium prausnitzii among children with obesity.⁶³

Lastly, the only longitudinal, observational study among school-aged children found minor differences in the microbial communities and no differences in diversity between normal weight controls and children with obesity.⁶¹ Bacteroides massiliensis and Bacteroides plebeius were enriched in the stool of children with obesity.⁶¹ This study⁶¹ was of good study quality and controlled for multiple comparisons. Despite some similarities in microbial taxa across the studies, there is a lack of consensus on specific gut microbes that are unique to school-aged children with overweight or obesity.

Adolescence (> 11 but < 18 years).

Of the included studies, 36% (n=12) evaluated associations between gut microbial composition and adiposity among adolescents. All studies were cross-sectional with substantial variation in results. Figure 2D provides a summary of the key microbial relationships with adiposity for the 12 included articles that focused on adolescents.

Three^29,46,49 of the five studies reporting on F:B ratio found a significantly higher ratio in adolescents with overweight/obesity compared to normal-weight controls, while two reported no difference.^47,50 There was similar consensus regarding an increased abundance of Firmicutes with greater adiposity^31,44,49 with one study⁵¹ reporting significantly lower Firmicutes abundance. The majority (60%) of these studies were of fair study quality.^29,47,49 In contrast, results were more mixed regarding the abundance of Bacteroidetes,^31,44,49,51 Proteobacteria^31,44,49,51 and Actinobacteria^31,44,49,51 when comparing adolescents with obesity to healthy weight controls. The only phyla with consistent reports was Verrucomicrobia^31,49,51 which did not differ by weight status.

Several of the studies reported key differences in microbial composition at the family taxonomic level. Similar to the results among school-aged children, Clostridiaceae abundance was significantly higher among adolescents with obesity.^31,44 Ferrer et al. found higher quantities of Streptococcaceae among adolescents with obesity but this was the only study that reported on this family which was of fair study quality.⁴⁴ Mixed results were observed for other reported families including Bacteroidaceae,^44,47,49 Lachnospiraceae,^46,47,49 Rikenellaceae,^42,44,47 Ruminococcaceae,^44,47,49 and Prevotellaceae.^44,47,49

Evaluation of differences in bacterial composition at the genus level among adolescents also yielded conflicting results. Sutterella,^42,46 Prevotella,^42,47 and Faecalibacterium^42,47 abundance varied among studies. Three studies^46,47,49 reported a significantly lower Bacteroides abundance among adolescents with obesity. Interestingly, Bai et al. found greater Dialister among adolescents with obesity but abundance of this genus was lower among overweight adolescents.⁴² This study had high study quality and controlled for potential covariates and adjusted for multiple comparisons. Overall, these findings demonstrate that microbes from specific genera may not only be associated with adiposity but also with the degree of adiposity.

There were several studies that investigated differences in concentrations of specific bacterial species. Two studies reported greater abundance of F. prausnitzii^31,43 while two other studies reported lower abundance of B. vulgatus^29,48 among adolescents with obesity. For the remaining bacterial composition differences, there was either a lack of agreement between the studies or only one study that reported on a microbe.

Evaluation of Study Quality

Table 2 provides the overall quality scores and ratings for each study while Supplemental Table 3 provides the responses to each of the 14 items on the Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies. Of the 33 included studies, eight studies^{37,42,50,55,60,61,64,65} received an overall quality rating of ‘good’ with four of the articles^60,64–66 being longitudinal studies. The article that received a response of ‘yes’ to the majority of the 14 items was Kamng’ona et al. which followed infants from birth to 18 months of age.⁶⁵ The majority (n=23) of the studies had an overall quality rating of ‘fair’,^{29,31,36–40,42–44,45,48,49,51,53–57,61,62–64,} with only two studies^46,48 being rated ‘poor’.

Table 2.

Summary of Quality, Potential Bias, Analytical Methods, and Covariate/Multiple comparison Adjustments of the Included Studies

Source	Quality Assessment		RTI Potential Bias	Statistical Methods Used	Statistical Adjustments
	Overall Score	Quality Rating			Covariate	Multiple Comparisons
Bai et al. Pediatric Obesity. 2019;14(4):e12480.42	7	Good	Information, Reporting	PERMANOVA, Linear decomposition model, Multivariate analysis	Age, sex, antibiotic and probiotic use	Benjamini-Hochberg correction using a false discovery rate
Balamurugan R et al. British J of Nutrition. 2010;103:335–338.43	5	Fair	Performance	Mann-Whitney U-test	None	None
Bervoets L et al. Gut Pathogens. 2013;5:1–10.29	5	Fair		Independent samples t-test, Mann-Whitney U-test, Chi-square test, Multiple linear regression	Age, gender	None
Borgo F et al. Childhood Obesity. 2017;13(1):78–84.63	6	Fair	Performance, Information, Selection bias/confounding, Reporting	Mann Whitney U test, ANOVA	None	None
Chen et al. PeerJ. 2020;8:e8317.53	6	Fair	Performance	Independent t-test (two-tailed), Wilcoxon rank-sum test, and LEfSe	None	None
Craig SJC et al. Scientific Reports. 2018;8:1–14.60	9	Good	Selection bias/confounding, information	Pearson’s chi-squared test, Function-on-Scalar Linear models; Choi significance measures, Mann-Whitney U and Kruskal-Wallis tests, Multiple linear regression, FLAME, and LEfSe	Gender, antibiotic exposure, medication use, delivery mode, intervention assignment, gestational diabetes, gestational weight gain, maternal smoking, family income, and dietary categories at age 2 years	Bonferroni-corrected p-values
Da Silva et al. Childhood Obesity. 2020;16(3):204–210.54	6	Fair	Selection bias/confounding, Information, Performance	Chi-square test, Independent t-test (two-tailed), Wilcoxon rank-sum test	None	None
Del Chierico F et al. Frontiers in Microbiology. 2018;9:1–12.31	5	Fair	Information	ANOVA, least significant difference (LSD) test, Mann–Whitney U-test, receiver operating characteristic (ROC), discriminant analysis (DA), principal component analysis (PCA), Spearman’s correlations. LEfSe	None	False discovery rate correction of p-values
Ferrer et al. Environmental Microbiology. 2013;15:211–26.44	4	Fair	Precision, Information, Information	NA	None	None
Forbes et al. JAMA Pediatrics. 2018;172:e181161.64	8	Good	Selection bias/confounding, Information	Logistic regression, Chi-squared tests, Multivariable logistic regression, Kruskal-Wallis tests, Post-hoc Dunn tests	Birth mode, parity, gestational diabetes, infant sex, birth weight, and maternal/neonatal antibiotic exposure	False discovery rate correction of p-values
Hollister et al. Childhood Obesity. 2018;14(2):122–130.61	9	Good	Selection bias/confounding, Information	Independent t-test (two-tailed), Wilcoxon rank-sum test, ANCOVA	Baseline microbiota composition	Benjamini-Hochberg correction
Hou et al. BioMed Res Int. 2017;7585989:1–8.46	3	Poor	Selection bias/ confounding, Performance, Information	Principal coordinate analysis using QIIME, Spearman correlation coefficient, PERMANOVA, Wilcoxon’s rank sum test, LDA algorithm using LEfSe	None	QIIME default method for pairwise comparisons to reject null hypothesis; therefore, no adjustments were made
Hu et al. PLoS ONE. 2015;10(7):1–14.47	6	Fair	Selection bias/confounding, Performance, Reporting	Chi-square test, Student’s t-test, Pearson partial correlation coefficient, Mann-Whitney U-test,	Age, gender	False discovery rate correction of p-values
Huerta-Avila et al. Nutrients. 2019;11(6):1–14.55	7	Good		Chi-square test, one-way ANOVAs with Bonferroni correction, Multiple linear regression with path analysis, Principal components and factor analyses	Age, sex, first-degree family history of obesity, leisure time physical activity	Bonferroni post-test correction
Ignacio et al, Clin Microbial Infect. 2016;22:258.e1–258.e8.56	6	Fair	Selection bias/ confounding	Chi-square tests, Spearman correlation, Kruskal-Wallis test, Multiple and logistic regression	Age and gender	Dunn and Tukey correction
Kamng’ona et al. Scientific Reports;2019;9:12893.65	10	Good	Selection bias/ confounding	Multiple linear regression, Logistic regression, Repeated measures ANCOVA	Intervention group; child age on day of stool collection; maternal age, height, body mass index, parity, education, HIV status, and hemoglobin at enrollment; household assets, food security, source of drinking water, residential location and access to sanitary facility; season at time of stool sample collection; mode of delivery (vaginal or cesarean); site of delivery; and child sex	Benjamini-Hochberg correction using a false discovery rate (FDR) of 0.15
Karlsson et al. Obesity. 2012;20(11):2257–61.57	5	Fair	Selection bias/ confounding, Information	Student’s t-test, Mann-Whitney rank sum test, Multivariate data analysis with principal component analysis	None	None
Karvonen et al. International J of Obesity. 2019;43:713–723.45	6	Fair	Selection bias/ confounding, Information	Chi-square tests, Mann-Whitney U test, Spearman correlation, Logistic regression	Maternal education Other confounders were examined but did not remain in final models because they did not change the estimates.	Bonferroni-corrected p-values
Leong et al. Am J of Clinical Nutrition. 2020;111(1):70–78.58	5	Fair	Selection bias/confounding	Student’s t-test, Chi-square test, Principal component analysis, Spearman correlation, Multivariate models	Model 1: Parity, household deprivation, mode of delivery. Model 2: breastfeeding, protein, fiver, non-starch polysaccharide, nut/seeds/legumes, meat/fish/poultry, BMI z-score	None
López-Contreras et al. Pediatric Obesity. 2018;13(6):381–388.59	6	Fair		Spearman correlation, Mann-Whitney U-test, Chi-squared test with Haldane’s correction	Age, gender	Benjamini-Hochberg correction using a false discovery rate (FDR) of 0.05, Haldane’s correction
Mansour et al. Res J Pharm, Biol, & Chem Sci. 2016;7:436–445.48	3	Poor	Selection bias/confounding, Information, Performance, Precision	NA	None	None
Méndez-Salazar et al. Frontiers in Microbiology. 2018;9(Oct):1–11.35	5	Fair	Precision, Information	Spearman correlation, One-way ANOVA, Kruskal-Wallis test, LefSe, Principal components analysis	None	Bonferroni-corrected p-values; Tukey’s adjusted p-values for multiple comparisons
Moran-Ramos et al. Gut Microbes. 2020; Jan 23:1–18.36	5	Fair	Selection bias/confounding	PERMANOVA, Canonical correspondence analysis, Spearman rank correlation, Multivariate logistic regression analysis using stepwise modeling	Age, sex, and monthly income	Benjamini-Hochberg correction using a false discovery rate (FDR)
Mousavi et al. Iranian Red Crescent Med J. 2018;20(4):1–6.37	7	Good	Reporting	Student’s t-test, One-way ANOVA, Kruskal-Wallis test, Mann-Whitney U-test, Chi-square test, Spearman’s rank correlation coefficient	None	None
Murugesan et al. Euro J Clin Micro & Infect Dis. 2015;34:1337–46.38	6	Fair	Selection bias/confounding, Performance, Information, Overall believability	Chi-square test, ANOVA, Kruskal-Wallis test, Principal components analysis	None	None
Nirmalkar et al. Nutrients. 2018;10(2009):1–23.39	5	Fair		Multivariate association with linear models, Pearson correlations, Spearman correlations, One-way ANOVA, Mann-Whitney U-test	None	Benjamini-Hochberg correction using a false discovery rate (FDR), QIIME default corrections; multiple testing was performed using the p.adjust function in R
Orbe-Orihuela et al. Salud Publica de Mexico. 2018;60(1):5–11.40	5	Fair		Kruskal-Wallis test, Logistic regression, Spearman rank correlation coefficient	Age, gender, total energy intake, and heredo-familial history of obesity	None
Payne et al. Nutrition & Diabetes. 2011;1(7):e12.41	5	Fair	Selection bias/confounding, Performance, Precision	One-way ANOVA	None	Tukey–Kramer HSD test
Riva et al. Environmental Microbiology. 2017;19(1):95–105.49	6	Fair		Student’s t-test, Chi-square test, Pearson correlation coefficient, Linear regression models, PERMANOVA	None	False discovery rate correction of p-values
Whisner et al. BMC Microbiology. 2018;18(1)1:11.50	7	Good		Wilcoxon-Kruskal Wallis test, Principal coordinate analysis, Linear discriminant analysis using LEfSe	None	False discovery rate correction of p-values
Xu et al. BMC Microbiology. 2012;12(1):283–289.62	5	Fair	Information	Chi-square test, Spearman’s correlation, Kruskal-Wall with Mann-Whitney U-test, One-way ANOVA	None	Bonferroni-corrected p-values
Yuan et al. J of Gastro & Hepat. 2014;29:1292–8.51	6	Fair		One-way ANOVA, Kruskal-Wallis test. Fisher’s exact test, Spearman’s rank correlation coefficient	None	Tukey’s correction for pairwise comparisons; Dunn’s multiple comparison tests
Zhao et al. Frontiers in Pediatrics. 2019;7:518.52	5	Fair	Selection bias/confounding, Information, Performance,	Independent samples t-test, Chi-square test, Kruskal-Wallis test, Wilcoxon rank sum test	None	None

The quality assessment tool for observational cohort and cross-sectional studies by the National Heart, Lung, and Blood Institute was used for quality assessment (https://www.nhlbi.nih.gov/health-topics/study-quality-assessment-tools). A quality rating algorithm was created where studies that had an answer of “yes” for more than 7 of the questions received a quality rating of “good”, studies that received an answer of “yes” for 4 to 7 of the questions received a quality rating of “fair” and studies that had an answer of “yes” for 3 or fewer of the questions received a quality rating of “poor”. Types of bias included selection, information, measurement, and confounding.

Evaluation of Potential Bias and Precision

There were 52 potential sources of bias or threats to precision identified among the included studies using the RTI 29-item bank. The most common types of bias were information (30%, n=16) and selection bias/confounding (32%, n=17). Eight studies^{29,39,40,49–51,55,59} had no potential sources of bias or threats to precision. Of the remaining studies, 21% (n=7) had one,^{31,36,37,43,53,58,62} 27% (n=9) had two,^{35,42,45,56,57,60,61,64,65} 18% (n=6) had three,^{41,44,46,47,52,54} and 9% (n=3) had four^38,48,63 potential sources of bias or threats to precision. Additionally, we report the various statistical methods used and if any multiple comparison or covariate adjustments were made. This information is summarized in Table 2 along with the overall quality assessment data.

Dietary Influence on Adiposity and Gut Microbial Modification

Despite the limited data available in children, there is a growing body of literature pertaining to the role that diet plays in microbial diversity and abundance. Growth of many beneficial microbes are inhibited by a Western-style diet that is high in fat. Conversely, the abundance of beneficial microbes is enhanced with plant-based diets including a variety of grains, fruits, and vegetables. Despite the known role that diet plays in the development of adiposity as well as the impact it has on the gut microbiome, 45.5% (n=15) of the included studies did not assess dietary intake. Among the studies that did evaluate diet (n=18), five used a dietary recall,^{35,38,39,43,50} six used a food frequency questionnaire^{36,49,55,59,63} or block screener,⁶¹ four used parent report of feeding practices or dietary intake,^45,58,60,64 two used non-validated dietary questionnaires,^41,42 and one used a five day food record.²⁹ Although almost half of the included studies evaluated diet, only 12% (n=4) of the included studies^35,42,45,58 evaluated diet-microbiota associations in relation to adiposity. Therefore, it was challenging to identify direct relationships among dietary factors, the microbiota, and adiposity outcomes. Table 3 provides a summary of the 26 microbes reported in four or more of the included articles, their metabolic function, and evidence-based dietary approaches reported in the included studies and broader literature that may modify their abundance.

Table 3.

Evidence-Based Dietary Approaches that Promote or Inhibit Obesity-Associated Gut Microbes

	Microbe	Metabolic Function	Diet-Microbe Relationships Reported in Included Studies	Diet-Microbe Relationships Reported in the Literature
				Increased Abundance	Decreased Abundance
Phylum	Actinobacteria	Degradation and decomposition of organic substances such as cellulose, polysaccharides, protein, fats, and organic acids; Production of bioactive enzymes that degrade starch.95	Not applicable	High-fat, high animal protein diet;96 Dairy, especially cheese and cheese flavoring;95 Light or low-calorie beer;95 Low-fiber diet26	Increase dietary fiber intake; reduce consumption of animal protein
	Bacteroidetes	Fermentation of polysaccharides and indigestible carbohydrates; Production of short-chain fatty acids.97	Not applicable	Increase fruit and vegetable consumption;98,99 increase dietary fiber intake;98,99 reduce consumption of animal protein	High-fat, high animal protein diet;100 Low-fiber diet100
	Firmicutes	Carbohydrate metabolism101	Not applicable	Dietary intake higher in fruits, vegetables, and fibers;102,103 High carbohydrate foods96,104	Low-fat diet
	Firmicutes:Bacteroidetes	Not Applicable	Positively associated with vegetable consumption and negatively associated with meat consumption60	See Firmicutes and Bacteroidetes above.	See Firmicutes and Bacteroidetes above.
	Proteobacteria	Anaerobic environment homeostasis;105 Colonization of mucus layer105	Not applicable	Insoluble fiber,106 plant-based foods (i.e. berries);106 Wild-gamey animal meat;106 Breast milk107	Mediterranean diet,106 Digestible starch and sugar106
	Tenericutes	Fermentation of sugars;108 starch synthesis and degradation;108 Acetate production;108 Cardiolipin and peptidoglycan synthesis108	Not applicable	Cereals and starchy foods;109 Fruit and vegetable consumption;109 Meat, honey, berries and tubers;106 Dietary fiber intake50	Italian-style gluten-free diet110
	Verrucomicrobia	Propionate producing and mucin-degrading;111 Decreased metabolic endotoxemia;102,112 Maintenance of intestinal integrity;113 Modulation of mucus thickness;114 Pro-inflammatory activity115	Not applicable	Foods high in polyphenols;114 Fermented foods such as unsweetened yogurt;116 Inulin and pomegranate extract;86 Vegetarian and vegan diets117,118	Western high-fat diet;114 Low FODMAP diets;114 Walnut consumption119
Family	Bacteroidaceae	Production of lactate and short chain fatty acids;108,109 Utilization of human milk oligosaccharides108,109	Not applicable	Increase consumption of fruits and vegetables and other plant-based foods;98 High-fiber plant-based foods98	High-fat diet;98 Diet rich in starchy foods such as potatoes98
	Christensenellaceae	Butanoate and propanoate metabolism;122 Butyrate synthesis122	Not applicable	Mediterranean diet;80 Plant-based foods such as vegetables proteins, dietary fiber, and polyphenols.123
	Coriobacteriaceae	Lipid metabolism124	Not applicable	Formula-fed infants;124 Moderate protein and carbohydrate intake125	High protein, low carbohydrate diets125
	Enterobacteriaceae	Sulfate-reducing126 Lactic acid production75	Not applicable	Moderate dietary protein restriction;91 breastfed infants;75,126 Animal meat;105,127 High-fat, low-fiber diets;117 Reduction in polysaccharide intake;25 Western-diet98	Early introduction of solid foods;70 Inulin;105
	Lachnospiraceae	Degrade complex polysaccharides to short-chain fatty acids for energy harvesting.112	Increased with omnivore diets42	Diets high in animal sources;93 Omnivore diet;93	High-fiber diet96,113
	Prevotellaceae	Degradation of dietary carbohydrates and proteins;110 Production of acetate, formate, propionate, and succinate.110	Not applicable	Animal protein, honey, and berries;106 Cereals and starchy foods;114 Legumes114	Plant-based and fermented foods;98
	Rikenellaceae	Nitrogen and protein metabolism;91 Butyrate synthesis122	Not applicable	High-fat diet;96,115 Milk and milk derivatives;114 Fruit juice;114 Fish;81 Refined carbohydrates.81	Decreased consumption of vegetables;81 Increased intake of legumes, nuts, and seeds.81
	Ruminococcaceae	Short-chain fatty acid production116	Not applicable	Mediterranean diet,106 Digestible starch and sugar106	Insoluble fiber,106 plant-based foods (i.e. berries);106 Wild-gamey animal meat;106
	Veillonellaceae	Propionate production;134 Lactate utilization;75 Degradation of fiber;106 Amino acid, carbohydrate, and energy metabolism135	Not applicable	Breastfed infants;75 Fiber intake;75 Plant-based diets;106 Carbohydrate intake110	Introduction of complementary foods;107 Gluten-free diet;110,129 Higher protein and energy intake110
Genus	Bacteroides	Degradation of dietary soluble polysaccharides;119,120 Utilization of amino acids for energy;119 Production of short-chain fatty acids and organic acids.119	Increased with omnivore diets;42 Positive association with fat and carbohydrate intake35	Higher consumption of animal protein and saturated fat;25,92 Low dietary intake of carbohydrates.92	Legumes;98 Iron and fiber from plant-based foods98
	Bifidobacterium	Maintenance of gut barrier, prevention of permeability and bacterial translocation;112 Metabolite production such as acetic acid70	Increased with vegetarian diets;42 Negative association with intake of nuts, seeds, and legumes;58 Positive association with fiber and non-starch polysaccharide58	Oligofructose;139 Probiotics;28,140 Yogurt consumption;141 Prebiotics;112,142 Breastfeeding;70,143 Whole grain consumption98,144	Gluten-free diet;145 Walnut consumption;119 Wild/gamey meat and foraging foods such as berries131
	Blautia	Sugar fermentation, flavonoid metabolism, leptin production, de novo lipogenesis and cholesterol genesis, protection against lipid accumulation146	Not applicable	High protein, low carbohydrate diet;125 Brown rice and whole grain barley;144 Savory snacks;140 Processed meat and milk consumption140	Reduced duration of exclusive breastfeeding;75 High fat diet;117 Low dietary fiber intake;80 Fish, eggs, and bean consumption140
	Dialister	Glucose metabolism;121 Soluble corn fiber fermentation122	Not applicable	Whole-grain bread;123 Increased consumption of B vitamins;124 Walnut consumption125	High protein, low carbohydrate diet126
	Faecalibacterium	Short chain fatty acid production, especially butyrate25,150	Increased with vegetarian diets;42 Positive association with nuts, seeds, legumes and meat, fish and poultry consumption58	High protein, low carbohydrate;125 high intake of folate and B vitamins149	Walnut consumption;119 Wild/gamey meat and foraging foods such as berries131
	Lactobacillus	Carbohydrate metabolism;127 Formation of lactic acid end products127	Increased with vegetarian diets42	Plant-based foods;127,128 Fermented foods (e.g. yogurt, cheese, pickles)127,129	Rice consumption;128 Fast food;130 Snacks and junk food130
	Prevotella	Degradation of dietary soluble polysaccharides;119,120 Production of acetate, formate, propionate, and succinate;110	Positively associated with monounsaturated fat and vitamin B135	Vegetarian diet;68 High-fiber diets;28,68 Mediterranean diet;25,132 High intake of carbohydrates26,133	Low dietary intake of sugars;26 Gluten-free diet25,134
	Sutterella	Potential pathogenic microbe leading to GI and metabolic dysfunction154,155	Not applicable	Cereals and starchy foods;109 Vegetables and fruits109	Soluble corn fiber148
Species	Bacteroides ovatus	Degradation of plant polysaccharides;138 Utilization of plant and host glycans;138 Degradation of polyphenols102	Not applicable	Human milk oligosaccharides;121 Pectin;117 Dietary polyphenols102	Mediterranean diet80
	Faecalibacterium prausnitzii	Fermentation of carbohydrates;37 Anti-inflammatory effects;141	Not applicable	High-fiber intake;92 Vegetarian and/or vegan diet;93	Low fiber intake;142 High insoluble fiber intake142

DISCUSSION

To the best of our knowledge, the present review is the first to systematically explore and critically evaluate the current state of the science regarding differences in gut microbiome composition among infants, children, and adolescents with overweight/obesity compared to normal-weight counterparts. Additionally, this review provides evidence-based dietary approaches that have been shown to promote or reduce the abundance of adiposity-associated microbes. Further, we provide best practice recommendations for future research evaluating adiposity-associated microbes during these formative years. Taken together, these studies demonstrate differences in early life microbiota composition that are associated with unique growth and dietary patterns that may increase the risk of overweight and obesity later in life. However, conflicting results among the included studies make consensus challenging.

Infancy/Toddlerhood

There is substantial evidence to support the dynamic relationship between feeding practices during infancy and toddlerhood and the early development of the microbiome. The protective effect of breastfeeding against obesity⁶⁷ may be due to key differences in the abundance of Bacteroidetes, Firmicutes, and Verrucomicrobia,⁶⁸ when compared to formula-fed infants. Interestingly, a decreased abundance of Verrucomicrobia was shown to be associated with overweight/obesity in toddlers, which was not affected by early life feeding practices (i.e. breastfeeding vs. formula-feeding) despite 31% of infants in this study consuming formula in the first year of life.⁴⁵ This relationship has been corroborated in adults with obesity.⁶⁹ The infant gut microbiome of formula-fed infants is more diverse with greater abundance of Staphylococci,⁷⁰ Bacteroides,^71,72 and Atopobium.^70,72 In the current review, there was no difference in Bacteroides,⁴⁵ but Atopobium parvulum⁶⁵ abundance was reduced among infants/toddlers with greater adiposity. One study found that formula feeding was associated with enrichment of Lachnospiraceae at 3–4 months which was subsequently enriched among overweight infants at 12 months,⁶⁴ indicating a unique microbial family linked to both early life feeding practices and adiposity. However, Lachnospiraceae enrichment was not associated with overweight/obesity among toddlers in another study.⁴⁵ Moreover, the quality of the studies included in this section received a higher quality rating of either fair or good with all of the studies controlling for covariates and adjusting for multiple comparisons. Taken together, further investigation of these potential microbe-adiposity relationships with breast vs. formula-feeding is needed.

This early developmental time period is also marked by the transition from semi-solid to solid foods. The introduction of complementary foods results in greater within-sample (alpha) diversity and a shift to Firmicutes and Bacteroidetes as the dominant phyla in the infant microbiota.⁷³ Of the included studies, there were no differences in microbial abundance among infants/toddlers by adiposity status for Bacteroidetes,^45,64 Firmicutes,^45,64 or F:B ratio.^60,64 This might be because the studies spanned birth to 36 months or the low alpha diversity of the infant microbiome.⁷⁴ The relative abundance of Lachnospiraceae and Ruminococcaceae have been noted to increase with the introduction of solid foods⁷⁵ and their enrichment at 3–4 months has been associated with greater adiposity at 12 months.⁶⁴ Lending further support for these microbial families, greater abundance of Dorea (family Lachnopiraceae)⁴⁵ and Ruminococcus obeum (family Ruminococcaceae)⁶⁵ have been observed among infants with overweight/obesity. Although studies included in this review did not explore specific feeding practices, Lachnospiraceae abundance may be an important indicator of diet-microbe-weight interactions as this family has been positively associated with protein intake among infants/toddlers.⁷⁵ Further, many of these findings are supported in adult populations,^69,76,77 indicating that microbial abundance of these microbial taxa may serve as early life indicators of obesity that continue into adulthood.

School-aged Children.

Unlike the studies among infants/toddlers, the current review found that school-aged children with overweight/obesity had significantly greater abundance of Clostridium^39,56 with lower relative abundance of Akkermansia muciniphila,^57,63 Oscillospira,^38,53 and Rikenellaceae.^53,59 A recent systematic review found that specific bacterial species belonging to the genus Clostridium were associated with adiposity;⁷⁸ however, studies among school-aged children in the current review did not find a relationship between Clostridium spp. and adiposity. Depletion of Oscillospira and Rikenellaceae have been reported among adults with obesity,⁷⁹ and associated with high-protein diets,⁸⁰ and legume consumption.⁸¹ There is strong evidence that lower abundance of Akkermansia muciniphila is associated with obesity in adults.⁸² Weight loss,^83,84 caloric restriction,⁸⁵ and dietary intake of pomegranate extract,⁸⁶ inulin (i.e. soluble fiber),⁸⁷ and fermentable oligo-, di-, and mono-saccharides⁸⁸ have all been reported to increase the relative abundance of Akkermansia muciniphila. Reduced abundance of B. plebeius was reported to be an obesity-associated microbe in two child studies^59,61 but dietary factors were not associated.⁵⁹ Overall, these results provide support for adiposity-associated microbial communities; however, differences in the above-mentioned microbial taxa may be heavily influenced by dietary changes during the formative years.

Unfortunately, the majority of studies included for this age range did not report on how diet may have contributed to findings regarding gut microbial abundance and measures of adiposity. There was only one study that examined diet-microbe-adiposity interactions. Leong et al found increased breastfeeding and lower BMI z-score were associated with reduced abundance of Christensenellaceae and Ruminococcaceae.⁵⁸ This study also found that less consumption of nuts, seeds, and legumes was associated with greater Bifidobacterium but reduced Bacteroides abundance. Additionally, it was reported that greater total fiber intake was associated with greater abundance of Faecalibacterium, Eubacterium, and Roseburia.⁵⁸ Méndez-Salazar and colleagues found that Bacteroides positively correlated with fat and carbohydrate intake, Proteobacteria was positively associated with fat intake, and Lachnospiraceae was negatively associated with total energy intake.³⁵ Unfortunately, this study did not extend its analyses to examine how these diet-microbe associations related to adiposity outcomes.

Adolescents.

Included studies of adolescents found conflicting results regarding differences in Firmicutes^31,44,49,51 and Bacteroidetes^31,44,49,51 phyla; however, two studies reported significantly higher abundance of Actinobacteria^31,44 among adolescents with obesity. This is somewhat surprising as adult studies have reported no difference in Actinobacteria between normal-weight adults and those with obesity.⁷⁶ Although these two studies did not explore dietary effects, previous research has shown that the abundance of Actinobacteria increases with high-fat diets²⁶ and decreases with vegetarian diets⁴² providing a potential dietary explanation for this microbe-adiposity relationship. Significantly lower abundance of Christensenellaceae^36,44 and Rikenellaceae^36,42,44,47 and greater abundance of Clostridiaceae^31,44 and Veillonellaceae^36,47 were observed in adolescent studies included in this systematic review. Previous research has found similar results among adults with obesity for Christensenellaceae,⁸⁹ Rikenellaceae,⁹⁰ and Veillonellaceae.⁹⁰ None of the 12 studies among adolescent populations examined diet-microbe-adiposity relationships, despite five studies^{29,42,43,49,50} assessing dietary factors. Two of these studies^43,49 evaluated differences in diet between children with and without adiposity. The remaining three studies^29,42,50 that evaluated diet-microbe relationships found no relationship between these bacterial families and dietary factors. The remaining studies did not evaluate dietary information which made it difficult to identify direct links between diet-microbe associations and adiposity. Nonetheless, previous research has reported reduced abundance of Rikenellaceae with decreased protein intake⁹¹ and consumption of vegetables.⁸¹ Similarly, Clostridiaceae abundance was reduced with inulin and polyphenol-rich pomegranate extract consumption.⁸⁶ Lastly, B. vulgatus^29,48 was significantly lower and F. prausnitzii,^31,43 was significantly higher among adolescents with obesity, but dietary effects on these microbe-adiposity relationships were not explored. Increased abundance of F. prausnitzii has been reported with diets high in fiber⁹² or vegetarian/vegan diets⁹³ whereas B. vulgatus abundance was reduced with greater rice consumption.⁹⁴ Exploration of the effects of dietary practices on adiposity-microbiome relationships are greatly needed in adolescent populations.

Strengths and Limitations

Although this review adds to the current body of literature about gut microbiota associations with adiposity early in life, it is not without limitations. First, most of the included studies received a quality rating of fair and primarily used cross-sectional study designs which only provide a snapshot of explored associations and cannot indicate cause-effect relationships. Furthermore, the heterogeneity among the bacterial communities appraised in the included articles made synthesis across studies challenging. Additionally, the included studies used a variety of methods for both adiposity and microbial analyses and obtained measures from participants living in various geographical locations, all of which may partially explain the inconsistencies in study results. Lastly, while 18 of the included studies collected dietary intake information, only 12% (n=4) of the included studies^35,42,45,58 evaluated both microbial abundance and diversity as well as dietary factors in order to assess the effect of diet-microbiota associations on adiposity. Due to a paucity in the literature, the current review referenced other articles that evaluated specific relationships between dietary factors and the gut microbiome to identify potential areas for intervention, but these findings should be confirmed in the age groups presented in this review.

CONCLUSION

Current data suggest the importance of the gut microbiome in the formative years on influencing obesity risk but identifying specific microbes that associate with obesity remains difficult given limitations in evidence. It is important that future research investigates the microbial community at different taxonomic levels, especially specific bacterial species, in relation to overweight and obesity in order to obtain a clear picture of which microbial strains serve as early indicators of obesity in infancy, childhood, and adolescence. This should be done for specific age groups as the microbiome differs across early life and may require unique dietary intervention depending on age. Strong evidence indicates the importance of breastfeeding for infants which impacts both microbiome community structure and weight outcomes. For children and adolescents, a well-balanced diet that consists of high intakes of fruits and vegetables with moderate animal fat and protein consumption seems to be associated with a healthy gut microbiome. This recommended dietary approach is not novel but remains difficult for most of the population considering the prominence of the Western-style diet. Overall, there is great need for longitudinal studies in infancy, childhood, and adolescence to understand how dietary behaviors and the gut microbiome interact to influence adiposity. Table 4 provides a summary of best practices for future research evaluating microbial relationships with adiposity, particularly the need to conduct formal mediation/moderation analyses that assess the influence of other factors (i.e. dietary factors) on microbe-adiposity relationships.

Table 4.

Best Practices for Future Research Evaluating Microbial Relationships with Adiposity

Area of Research	Best Practice Suggestion
Population	Avoid spanning across multiple developmental periods. Instead focus on one age group to capture unique changes in microbes and adiposity risk in relation to that age.
	If the study includes a wide range of ages, i.e. spans different periods of development, report the microbial relationships separately for different age groups.
Study design	More longitudinal studies are needed particularly when assessing microbe-diet associations and their impact on obesity.
	Multiple data collection timepoints throughout the various developmental time periods that include fecal, diet, and adiposity data collection are needed.
Microbial Analysis Methodology	Use the same PCR primers as previous research. Primers for the V4 region of the 16S rRNA gene is recommended by the Human Microbiome Project, when performing high-throughput 16S sequencing.
	Differentiate bacterial taxa at the species level whenever possible.
	If possible, consider using metagenomics shotgun sequencing in order to gather a more comprehensive list of microbial strains present in the sample.
Adiposity Measures	Use a variety of measures for assessing adiposity to ensure that there are measurement similarities across studies.
	Report all adiposity measures as part of the article or in supplemental materials.
Dietary Intake Measures	Evaluate dietary intake using 24-hour dietary recalls.
	Assess diet on the same day as fecal collection whenever possible.
	Control for diet in analyses through mediation/moderation analyses or multivariate models.
Statistical analyses	Control for multiple comparisons by using available programs such as QIIME2.
	Formal mediation or moderation analyses to examine the effect of other factors (diet, antibiotics, geography) on microbe-adiposity relationships.
Covariates/Confounders	Report key information such as geographic location, time period of the study, season of fecal collection, birth mode, antibiotic use, etc., which may also interact with variables of interest to influence the microbiome.

These best practices are only suggestions based on the limitations of the current systematic review and those reported in a recent publication (See Reference). We suggest this article as a good reference for developing research projects that study, analyze, and interpret the human gut microbiome. Reference: Allaband C et al. Clinical Gastroenterology and Hepatology. 2019;17(2):218–230.

Supplementary Material

Sup Info 2

Supplemental Table 1. Study Characteristics and Evaluation of Included Articles

Supplemental Table 2. Microbial Relationships with Adiposity Reported among Two or More of the Included Studies

Supplemental Table 3. Complete Evaluation of Quality and Potential Bias of Included Studies