Searching for a Consensus Among Inflammatory Bowel Disease Studies: A Systematic Meta-Analysis

Abstract

Background: Numerous studies have examined the gut microbial ecology of patients with Crohn's disease (CD) and ulcerative colitis, but inflammatory bowel disease-associated taxa and ecological effect sizes are not consistent between studies.

Methods: We systematically searched PubMed and Google Scholar and performed a meta-analysis of 13 studies to analyze how variables such as sample type (stool, biopsy, and lavage) affect results in inflammatory bowel disease gut microbiome studies, using uniform bioinformatic methods for all primary data.

Results: Reduced alpha diversity was a consistent feature of both CD and ulcerative colitis but was more pronounced in CD. Disease contributed significantly variation in beta diversity in most studies, but effect size varied, and the effect of sample type was greater than the effect of disease. Fusobacterium was the genus most consistently associated with CD, but disease-associated genera were mostly inconsistent between studies. Stool studies had lower heterogeneity than biopsy studies, especially for CD.

Conclusions: Our results indicate that sample type variation is an important contributor to study variability that should be carefully considered during study design, and stool is likely superior to biopsy for CD studies due to its lower heterogeneity.

Keywords: CD; IBD; UC; alpha diversity; beta diversity; meta-analysis; microbiome; sequencing.

Figure 1.

PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) diagram and bias funnel plot. A, PRISMA flow chart of study selection process. B, Bias funnel plot of odds ratios for Crohn’s disease (CD) samples to have below-average observed richness. Egger’s regression test P > .05 C, Bias funnel plot of odds ratios for ulcerative colitis (UC) samples to have below-average observed richness. Egger’s regression test P > .05. We acknowledge the work of all authors of each study; however, we removed “et al” in figures for simplicity in visualization.

Figure 2.

Odds ratios (ORs) for below-average observed richness in the gut microbiome of inflammatory bowel diseasepatients: A, Crohn’s disease (CD) patients; and B, ulcerative colitis (UC) patients. We acknowledge the work of all authors of each study; however, we removed “et al” in figures for simplicity in visualization. CI, confidence interval.

Figure 3.

The relationship between odds ratios (ORs) for low richness and sample type: A, Crohn’s disease (CD) biopsy; B, CD stool; C, ulcerative colitis (UC) biopsy; D, UC stool. We acknowledge the work of all authors of each study; however, we removed “et al” in figures for simplicity in visualization. CI, confidence interval.

Figure 4.

Log2 fold changes in the abundance of genera in ulcerative colitis samples compared with control subjects. Ulcerative colitis vs non–inflammatory bowel disease. The y-axis colors pastel green, gray, magenta, orange, black, and lime green indicate phyla Actinobacteria, Bacteroidetes, Firmicutes, Fusobacteria, Proteobacteria, and Verrucomicrobia, respectively. *, **, and *** denote P < .05, P < .01, and P < .001, respectively. Genera are organized by family within phyla (See online version for color figure)..

Figure 5.

Log2 fold changes in the abundance of genera in Crohn’s disease samples compared with control subjects. Crohn’s disease vs non–inflammatory bowel disease. The y-axis colors green, gray, blue, magenta, orange, and black indicate phyla Actinobacteria, Bacteroidetes, Euryarchaeota, Firmicutes, Fusobacteria, and Proteobacteria, respectively. *, **, and *** denote P < .05, P < .01, and P < .001, respectively. Genera are organized by family within phyla (See online version for color figure).

Figure 6.

Inflammatory bowel disease–associated taxa: interstudy agreement. All disease-associated taxa identified by DESeq2 (Q < .2, Benjamini-Hochberg false discovery correction) in at least 2 (A) Crohn’s disease (CD) or (B) ulcerative colitis (UC) studies are shown. Color corresponds to direction of inflammatory bowel disease–associated change. C, Total numbers of CD- and UC-associated genera in at least 2 studies (See online version for color figure).