Antibiotics Associated With Increased Risk of New-onset Crohn's Disease but not Ulcerative Colitis

A Meta-analysis

Ryan Ungaro MD; Charles N Bernstein MD; Richard Gearry MB ChB; PhD; Anders Hviid MSc; Kaija-Leena Kolho MD; PhD; Matthew P Kronman MD; MSCE; Souradet Shaw MSc; Herbert Van Kruiningen MD; Jean-Frédéric Colombel MD; PhD; Ashish Atreja MD; MPH


Am J Gastroenterol. 2014;109(11):1728-1738. 

In This Article


The results of this meta-analysis suggest that exposure to antibiotics increases the risk of new-onset IBD. When stratifying by type of IBD, antibiotic exposure was associated with an increased risk of developing CD but not UC. We found that the magnitude of risk of new CD is greater for children than for adults. All classes of antibiotics studied, with the exception of penicillins, were associated with new diagnoses of IBD. Interestingly, the types of antibiotics showing the strongest association were fluoroquinolones and metronidazole.

Although it is impossible to draw causal links on the basis of these data, there are some possible implications and explanations for our findings. First, our findings may support the importance of disruptions in the microbiome in the pathogenesis of IBD. The link between antibiotic exposure and new IBD seems biologically plausible. It is known that the microbiome likely has an important role in the pathogenesis of IBD. Studies have shown a decrease in the diversity and stability of both mucosa-associated bacteria and fecal bacteria in patients with CD and UC.[5] For example, the largest cohort microbiome study to date recently found that newly diagnosed CD patients have increased Enterobacteriaceae, Pasteurellaceae, Veillonellaceae, and Fusobacteriaceae, and decreased Erysipelotrichales, Bacteroidales, and Clostridiales.[6]

Antibiotics have been shown to alter the composition of the human gut microbiota by decreasing taxonomic richness and diversity.[28] In addition, metaproteomic analysis has demonstrated that antibiotics decrease protein expression and negatively affect the overall metabolic status of the gut microbiome.[29] Within a recently reported cohort of newly diagnosed CD patients, a subset of patients were taking unspecified antibiotics. Antibiotics were noted to amplify the dysbiosis seen in CD patients.[6] Although the microbiome may recover to its initial state within days to weeks after antibiotic treatment, some studies have shown a longer-term impact of antibiotics on specific microbial populations that can persist for months to years.[28,30,31] Even if an antibiotic-induced dysbiosis does not last long, it is possible that a temporary alteration in the microbiome could prime or trigger dysregulated mucosal immune responses in susceptible individuals.

It is unclear as to why antibiotic exposure was associated with new-onset CD and not UC. Studies have suggested a difference in the microbiota between CD and UC patients.[5] For example, one study using mucosal biopsies found that CD patients had significantly altered bacterial flora compared with both healthy controls and those with UC with increased Proteobacteria and Bacteroidetes and decreased Clostridia.[32] Another study using both biopsies and stool samples found that the microbiome composition in UC patients was generally less dissimilar than healthy subjects, with more marked differences in CD.[33] Although speculative, it is possible that antibiotics are more likely to cause changes in the microbiome that predispose to CD as opposed to UC or that alterations in the microbiota are more of a driving factor in the development of CD than in UC. This effect may be accentuated in pediatric populations. Our finding that pediatric populations appear to have an increased association of antibiotic use with new-onset CD compared with adults may reflect the less stable nature of the microbiome earlier in life. During the first 3 years of life, the microbiome appears to undergo marked changes and significant maturation toward an adult-like composition with greater interpersonal variation.[34] It is possible that antibiotics may therefore have a greater impact during childhood when the gut microbiota composition is still developing.

Nearly all antibiotic types conferred an increased risk of new IBD in our analysis. The magnitude of this risk appears to vary among antibiotic classes. It is unclear as to what might account for these differences. One possibility may be the differing impacts on the microbiota. There are currently limited data on the effects of specific antibiotics on the composition and function of the gut microbiota. The two antibiotics that were most strongly associated with new-onset IBD were fluoroquinolones and metronidazole. This is somewhat paradoxical, as these antibiotics are commonly used in the treatment of IBD, although at times with mixed results.[35] Both of these medications are also frequently prescribed for gastrointestinal infections that, as described below, are possible risk factors for new-onset IBD. All three studies that reported data on fluoroquinolones and metronidazole excluded time between antibiotic exposure and new diagnosis of IBD, limiting the possibility that these antibiotics were prescribed for yet to be diagnosed IBD.[11,12,16]

Interestingly, there is some evidence to suggest that metronidazole and fluoroquinolone treatment may cause alterations in the gut microbiota that mirror changes observed in IBD patients. In addition to diminishing the intestinal mucus layer and goblet cell function, metronidazole treatment in the Citrobacter rodentium colitis mouse model appears to deplete Bacteroidales that has been shown to be decreased in CD patients.[36] In an in vitro model of common human anaerobes, metronidazole decreased Bacteroides (which is commonly diminished in CD) but increased Bifidobacteriaceae that tend to be depleted in CD patients.[37] An older study found that metronidazole can increase Escherichia coli in the feces.[38]E coli, in particular adherent-invasive strains, are increased in CD patients.[5] There is also evidence on the effects of fluoroquinolones on the human microbiome. A 5-day course of ciprofloxacin markedly decreases Bacteroides, Lachnospiraceae, and Ruminococcaceae in the stool of healthy subjects.[30] These bacteria have all been noted to be depleted in CD patients.[6] Another study looking at ciprofloxacin demonstrated anywhere from a 36 to 82% decrease in species diversity and the absence of Clostridiales taxon following treatment.[28]

An additional possibility is that instead of playing a direct role in the development of IBD, antibiotic exposure may serve as a surrogate marker for another risk factor for new-onset IBD. One possibility is that the association of antibiotics with new IBD reflects the previously suggested link between enteric infections and the development of IBD. For example, a study using the US General Practice Research Database also found an increased risk of IBD following episodes of gastroenteritis. The hazard ratio following an episode of gastroenteritis was 2.4 (95% CI 1.7–3.3), with the risk being greater for patients also treated with antibiotics.[7] Another study, a nested case–control study using data from the US military, demonstrated that gastroenteritis increased the risk of IBD with an OR of 1.40 (95% CI 1.19–1.66).[39] This risk was slightly higher for CD than for UC. When looking at specific infections, Salmonella and Campylobacter intestinal infections have been associated with an increased risk of new IBD in a Danish cohort.[40] However, these studies need to be viewed cautiously, as the association between new IBD and enteric infections may be significantly influenced by detection bias owing to increased rates of stool testing.[41]

Four of the studies included in this meta-analysis reported the indications for antibiotic use (Supplementary Table S4a Two of the studies only reported on the indications for metronidazole.[11,12] Shaw et al.[12] found that the majority of metronidazole use was for indications related to gastrointestinal diagnoses. However, in the study by Kronman et al.,[11] only 10.6% of metronidazole prescriptions were related to gastrointestinal indications. Most of the specified metronidazole indications were either vaginitis or oral infections. However, nearly 40% of the indications in this study were not specified. In Shaw et al.,[14] the vast majority of antibiotics were given for otitis media or respiratory infections. Margolis et al.[17] reported on antibiotics given for acne. If the majority of antibiotics were given for nongastrointestinal indications, this might suggest that antibiotics may be having a more direct role in increasing IBD risk (e.g., increasing intestinal dysbiosis) and less likely to have been administered for yet to be recognized IBD. However, the majority of studies in this meta-analysis did not report antibiotic indications, and thus ultimately we are unable to draw conclusions about the reasons for antibiotic use.

When looking at excluded studies that used infections as a surrogate for antibiotic use, the majority of infections involved the upper and lower respiratory tract (Supplementary Table S4b[9,22,24] These three studies looked at the association of childhood infections with IBD. One case–control study that analyzed otitis media, pharyngitis, tonsillitis, and scarlet fever found that CD patients were more likely to report childhood pharyngitis (OR 2.14, 95% CI 1.30–3.51), tonsillitis (OR 1.65, 95% CI 1.02–2.63), and colds (OR 2.42, 95% CI 1.38–4.24).[24] No association was seen with UC in this study. Another case–control study used a national database from Sweden and evaluated the association between CD and common childhood infections that typically require antibiotics from birth to the age of 5 years.[9] Pneumonia was significantly associated with both pediatric and adult-onset CD. The third study examined environmental exposures to evaluate the hygiene hypothesis in pediatric-onset CD.[22] Participants completed a questionnaire and were asked about "physician-diagnosed childhood infections," showing a nonstatistically significant trend toward increasing risk of CD (OR 1.4, 95% CI 0.9–2.5).

One hypothesis related to these previous studies is that the association between IBD and preceding infections (that then lead to antibiotic use) is evidence supporting a dysregulated immune response due to an infectious trigger or alternatively an underlying predisposition to infections in patients with IBD. If this hypothesis is true, our finding that antibiotics are associated with CD but not UC might reflect underlying differences in the immune systems of patients with CD vs. UC. Although it is not possible to make these conclusions based on the current data, it is an intriguing hypothesis given that many genes affecting innate immunity have been associated with IBD.[42]

Interestingly, IBD is not the only immune-mediated disease in which antibiotic exposure may be a risk factor. For example, antibiotic exposure in children may increase the risk of subsequently being diagnosed with asthma.[43] Early-life infections and antibiotic use have also been associated with atopic dermatitis.[44] A case–control study using a Swedish national database found that antibiotic exposure was associated with developing celiac disease.[45] Similar to the current study, noncausal explanations for these associations cannot be fully excluded.

There are several limitations to this study. First, it is difficult to exclude the possibility of protopathic bias in the included studies. Patients may have received antibiotics for symptoms of yet to be diagnosed IBD. However, 9 of the 11 included studies excluded some time frame before index IBD diagnoses. Although this does not eliminate protopathic bias, it should limit this bias. Second, a possible confounder that could affect our findings is that the prodromal symptoms of CD may differ from UC. Excluding time before diagnosis may not capture associations with UC, as it may be more likely to present more acutely. However, studies that looked at antibiotic exposures without an exclusion time frame in the primary analysis or as a secondary analysis did not find an association with UC.[10,11,17] Third, there was significant heterogeneity among the studies when all were grouped together. The heterogeneity appeared to be due to the age of the study population, predominant type of IBD in study population, and study design. Heterogeneity decreased when looking at subgroups stratified by age (children vs. adults), disease type (CD vs. UC), and study design (excluding survey studies). In addition, a majority of studies included in the analysis were retrospective. Four of the included studies were survey studies that may be subject to recall bias. However, excluding these studies in sensitivity analyses affected the magnitude but not the statistical significance of our findings. Many of the studies measured antibiotic exposure by prescription dispensation based on health-care system or pharmacy databases, and data confirming consumption of the medication were not available. It is therefore possible that some antibiotics were prescribed or dispensed by a pharmacy but not used. It is possible that the increased risk seen in children may be related to increased antibiotic exposure during childhood in general compared with adults. Furthermore, although we are unaware of quality data to suggest this, children may be less likely to undergo invasive procedures used to diagnose IBD, and therefore the diagnoses may not be as rigorously established and could overestimate the risk in children seen in this study. However, these considerations are unlikely to change the overall trend of increased risk of CD in both children and adults. Another limitation is that the included study populations were from Europe, Canada, or New Zealand. Our findings may therefore only hold significance in these geographic regions. Last, only publications that were written in English were included because of difficulty in accurately extracting data from studies in foreign language. However, no pertinent non-English language articles were noted in our search, and thus it is unlikely that significant articles were missed.

In conclusion, exposure to antibiotics, in particular in children, is associated with an increased risk of new-onset CD but not UC. This risk appears greatest with metronidazole and fluoroquinolones. Future prospective studies investigating environmental risk factors for IBD should include an exploration of previous antibiotic exposures. Although there are already myriad reasons to be judicious with antibiotic use, it may be prudent to limit antibiotic exposure unless it is absolutely necessary, particularly in patients at an increased risk for IBD (e.g., strong family history of IBD). Further research on how specific antibiotics affect the microbiome may provide valuable information for developing treatment strategies that involve modulating the intestinal flora.