Introduction

Celiac disease (CeD) is an immune-mediated inflammatory disorder manifested by an aberrant immune response to dietary gluten in genetically predisposed individuals, which can ultimately lead to intestinal malabsorption1. In addition to genetic and environmental factors, abnormalities of the gut microbiota have also been identified in CeD. Various pathogens, including bacteria, viruses, and fungi such as reovirus, human adenovirus serotype 2, rotavirus, Candida albicans, Clostridioides difficile, Bacteroides fragilis, Prevotella species, Lachnoanaerobaculum umeaense, Actinomyces graevenitzii, Neisseria flavescens, and Proteobacteria have been linked to developing CeD, presumably due to stimulating immune activation and disruption of the gut epithelial barrier2,3. Additionally the gut microbiota influences immune responses through mediators including short-chain fatty acids, immunoglobulin A, and toll-like receptors, regulating intestinal permeability and inflammation2. CeD has additionally been associated with distinct microbial profiles. Alterations in the microbiota have been observed in the oral, duodenal, and fecal microbiota, with an increased abundance of Proteobacteria, particularly Escherichia coli and Neisseria, and a decreased presence of Firmicutes (Lactobacillus, Streptococcus) and Actinobacteria (Bifidobacteria). Sequencing studies confirm that Proteobacteria are dominant in active CeD, with Neisseria flavescens significantly more abundant in active CeD patients than in those on a gluten-free diet4.

Gastrointestinal microbial overgrowth (MO) describes an overabundance of microorganisms (bacteria, fungi, archaea) within the small bowel which leads to gastrointestinal symptoms. MO can be further categorized by microtypes based on the causative organisms. These include small intestinal bacterial overgrowth (SIBO), intestinal methanogen overgrowth (IMO), and small intestinal fungal overgrowth (SIFO). SIBO has consistently been associated with CeD. A systematic review and meta-analysis of 14 studies (742 CeD patients) found that MO prevalence in CeD patients was 18.3%, with a significantly higher prevalence compared to controls (OR 5.1, 95% CI: 2.1–12.4, p = 0.0001). Breath tests detected a higher prevalence (20.8%) than culture-based methods (12.6%). Among CeD patients nonresponsive to a GFD, MO prevalence was not significantly higher than in those responsive to a GFD5. However, interpretation of these findings is challenging given the numerous controversies around the diagnosis of MO, including suboptimal sensitivity and specificity of diagnostic tests and evolving definitions of test positivity over time. Additionally, the meta-analysis did not assess the prevalence of IMO, a form of MO caused by overgrowth of methanogenic archaea and diagnosed on breath testing which has been associated with constipation6. Similarly, little is described about SIFO in CeD. Finally, it is not known whether SIBO correlates with CeD histologic activity.

The aim of this retrospective study is to provide a contemporary update on the associations of MO in CeD and ascertain whether SIBO is associated with CeD-related histologic changes.

Results

Intestinal aspirate cultures

A total of 256 patients with CeD underwent EGD with aspirate culture (Table 1). Of these, 55% underwent EGD within 6 months of diagnosis, 10.5% within 3 years, and 34.4% after 3 years.

Table 1 Demographic information of patients with CeD who underwent EGD with aspirate culture
Full size table

SIBO was positive in 17.6% at a threshold of ≥105 CFU/mL and 49.6% at a threshold of ≥103 CFU/mL. SIBO was significantly more common in refractory CeD, at 36.8% vs. 16% (p = 0.053) at ≥105 CFU/mL and 78.9% vs. 47.3% (p = 0.008) at ≥103 CFU/mL (Table 2). Eleven patients (4.3%) had ≥103 CFU/mL of fungi yeast or fungi; five (45.5%) of these were in conjunction with SIBO, while the remaining six (54.5%) were isolated SIFO without SIBO. The distribution of total microbiota, yeast recovery, and bacterial speciation are listed in Supplementary table 1.

Table 2 SIBO prevalence Stratified by Marsh score (≥2) and refractory CeD
Full size table

Marsh score distribution was 56.2% 0, 12.7% 1, 0.4% 2, 23.5% 3a, 5.6% 3b, and 1.6% 3c. There was no association between the Marsh score and SIBO result (Table 2).

Breath tests

Thirty-nine patients underwent breath testing (36 with glucose, 3 with lactulose), and nine (23.1%) had positive results. Among these nine positive tests, none were due to elevated hydrogen alone, six (66.7%) were due to elevated methane alone, and three (33.3%) were due to elevations in both hydrogen and methane. Therefore, 100% of positive breath tests featured elevated methane.

Seven patients underwent both testing modalities (Table 3). Median time between tests was 76 days (range 1–270). Among the two tests performed only days apart, one showed discrepant results, with a negative breath test but positive aspirate culture. Agreement between tests was 28.6%.

Table 3 Results of MO testing among patients who had undergone both breath testing and aspirate culture
Full size table

Discussion

MO is common among patients with CeD. SIBO was identified via intestinal aspirate culture at 17.6% by the ≥105 CFU/mL and 49.6% by the ≥ 103CFU/mL threshold. There was no association between the presence of SIBO and Marsh score; however, SIBO was significantly more common in refractory CeD than in non-refractory CeD at both thresholds, with rates of 36.8% and 78.9%, respectively. SIFO was rare (4.3%). MO was identified on breath testing in 23.1% of patients, due to mainly to IMO (100%).

This is the largest study to date assessing SIBO in CeD. All previous studies had smaller numbers, and great variability in SIBO prevalence, ranging from 2 to 50% when diagnosed via aspirate culture (Table 4)7,8,9,10,11,12. This variability may be in part due to differences in sampling techniques as well as patient populations, since studies include a wide range of patients including healed CeD, those, with active disease, and patients with refractory CeD. This is important because disease activity has been associated with the risk of SIBO. Indeed, in the prior largest study, 9.3% of CeD patients overall had SIBO (at ≥105 CFU/mL), but while the prevalence of SIBO was 11% both at the time of diagnosis and in those with non-responsive CeD, 0% of patients in remission had SIBO7. Our results may overestimate the prevalence of SIBO in CeD since our tertiary care population has an over-representation of complicated cases of CeD such as those with persistent symptoms, non-responsive CeD, or refractory CeD, in whom we found especially high rates of SIBO.

Table 4 Summary of previous studies assessing SIBO in CeD via aspirate culture
Full size table

We chose to analyze SIBO at two thresholds. While the historical threshold has been ≥105 CFU/mL, this has not been validated. Indeed, evidence from multiple studies has shown that a bacterial load of ≥105 CFU/mL is most commonly restricted to patients with a history of abdominal surgery (such as Roux-en-Y gastric bypass or Billroth II gastrectomy), and that healthy individuals rarely exceed ≥103 CFU/mL13. Furthermore, small bowel sequencing has linked a bacterial load of ≥103 CFU/mL with patient symptoms, characteristic dysbiosis, and metabolomic changes14. Based on this threshold, SIBO was very common in CeD, at 49.6% (and 78.9% in refractory CeD). This is considerably higher than a previous study at our institution which found a 20.8% prevalence at this threshold7.

Notably, the specific bacteria identified in our SIBO patients are consistent with the types of bacteria found in sequencing studies as causative of SIBO, including an overabundance of members of the family Enterobacteriaceae and the so-called “disruptor taxa” such as Klebsiella and Escherichia14,15. This is also in line with previous studies of the microbiota in patients with CeD, which has demonstrated a predominance of Enterobacteriaceae and specifically Klebsiella and Escherichia16,17.

SIBO was not more common among those with higher Marsh scores. This finding is in concordance with a historical study from our institution demonstrating no difference in the number of intraepithelial lymphocytes or the degree of histologic abnormalities between CeD patients with and without SIBO7. Such findings suggest that patients with more advanced CeD are not at higher risk of SIBO; therefore SIBO testing should be considered among CeD patients at any point in their disease with ongoing symptoms. This finding however is slightly limited by the fact that only 29.9% of patients had a Marsh score of 2 or higher, with the majority of patients demonstrating mucosal healing at the time of assessment. Notably, patients with refractory CeD did have significantly higher rates of SIBO at both thresholds, of 36.8% and 78.9% for the more and less stringent thresholds, respectively. Therefore SIBO may be particularly prevalent in patients with refractory disease.

SIFO was rare in our cohort, identified in only 4.3% of patients, and mainly co-existing with SIBO. However, this result must be interpreted with caution, since intestinal aspirate cultures at our institution are only grown on bacterial media (unless a concomitant specific fungal culture is also ordered). Fungi are only reported by the lab if there is a clear growth of fungi even on the bacterial media. Because of this, our results may underestimate the prevalence of SIFO in CeD. While there is no consensus on the significance, workup, and management of SIFO, fungal dysbiosis has been documented in CeD (in fact Candida albicans has even been hypothesized as a trigger for CeD)18, so future studies should seek to better understand the prevalence and significance of fungi in patients with CeD19.

Our cohort featured a much smaller number of patients who had undergone breath testing to diagnose MO. While only 39 patients were identified, we employed the most updated criteria to interpret breath test results20,21. The previous studies of MO in CeD are heterogeneous in terms of patient population, test substrate, criteria for hydrogen and methane (if assessed at all), and duration of the test (Table 5), and prevalence ranged 6.1–66.7%12,22,23,24,25,26,27,28,29. Similar to the previous studies utilizing aspirate cultures, the breath test studies vary considerably in terms of substrate, test protocol, and criteria for positive tests. Our results identified MO prevalence of 23.1%, with 100% representing IMO (mostly in isolation, with a minority also featuring hydrogen SIBO). The majority of tests (92.3%) utilized glucose as substrate. This is relevant because the lactulose breath test is severely limited by its susceptibility to false positive results due to orocecal transit time30,31. Patients with diarrheal illnesses have rapid orocecal transit; in these cases, it is impossible to distinguish a rise in breath hydrogen as resulting from SIBO or from colonic fermentation30. This may be particularly relevant in CeD, where MO testing is often ordered for diarrhea. However, unexpectedly, CeD has actually been associated with prolonged orocecal transit times32,33.

Table 5 Summary of previous studies assessing MO in CeD via breath testing
Full size table

It is also intriguing that all positive tests in CeD were due to methane (mostly alone, with a minority also featuring elevated hydrogen), since IMO has consistently been associated with constipation6 and slow intestinal and colonic motility34. This finding is in contrast to another small study which found that methanogens (such as Methanobrevibacter smithii) were absent in fecal samples from patients with CeD35. Possible explanations for IMO in CeD include prolonged transit time32,33 or the development of chronic constipation (due to insufficient dietary fiber or the development of a rectal evacuation disorder, the latter which is associated with MO, especially IMO36), which is common in CeD after starting the gluten-free diet37,38. Since the vast majority of patients in our cohort had undergone MO testing via aspirate culture (which does not readily identify methanogens), it is possible that we are underdiagnosing IMO in patients with CeD.

Finally, there were seven patients who underwent both MO testing modalities, including two who had both performed within three days of each other, and four within 90 days. Among the two patients who had both performed within three days of each other, one of them was discrepant, showing a negative breath test and a positive aspirate culture. It is challenging to interpret the remaining studies since sufficient time had elapsed for treatment to be taken or for the clinical situation to change. Prior studies evaluating both diagnostic modalities together have shown poor agreement between the two. In one study of 40 individuals, jejunal aspirate culture was positive in 35% while the lactulose breath test was positive in 45% and the glucose breath test in 30%. Furthermore, there was poor agreement between culture and the lactulose breath test39. In a study of 139 individuals, duodenal aspirate culture was positive in 44.6% and glucose breath test in 27.3% (diagnostic agreement of 65.5%). Poor agreement between tests may arise due to suboptimal sensitivity and specificity of breath testing which derives from the numerous confounders affecting breath gas dynamics30,31, or from false positive aspirate cultures which can occur from contamination of the oropharyngeal flora40. Future studies should prospectively evaluate which modality is most sensitive and specific for MO in CeD.

Our study does have some limitations. First, this was a retrospective study assessing MO among patients with CeD undergoing testing for MO. This introduces selection bias toward symptomatic patients, and does not provide a true prevalence estimate for MO among all patients with CeD. Related and as described prior, selection bias may also be introduced by nature of the unique population that our tertiary care referral center draws. Lacking from our analysis are the indications for MO testing, so we do now know why the tests were ordered and we cannot predict which symptom(s) are most suggestive. However, these tests were most likely ordered because the patient was symptomatic, and at our institution, the most common indication for testing is diarrhea. It is also worth noting that previous research has found that specific symptoms have limited ability to predict MO41. Additionally, we do not have data on response to MO treatment. Furthermore, this study did not control for potential confounders which may explain the development of MO such as proton pump inhibitor use and dysmotility or comorbidities such as inflammatory bowel disease or microscopic colitis. There is also concern that the aspirates cultures could be overestimating SIBO due to contamination by the oropharyngeal flora, since our institution does not employ a sterile double lumen catheter, and previous research from our system found that 20% of aspirate samples were contaminated, and culture and breath test results differed in 36.5% of subjects40. Similarly, breath test data were not assessed for confounders such as transit time, and breath test result was not assessed in relation to histology, since patients often underwent breath testing in the absence of recent endoscopic surveillance of their CeD. Despite these limitations, we believe that MO does represent a clinically significant finding in CeD. This is supported by previous studies which have specifically investigated the consequences of MO and its treatment in CeD. For example, in a prospective study of 15 patients with nonresponsive CeD, 10 had MO (diagnosed via lactulose-hydrogen breath test), and after treatment with rifaximin, 100% had complete symptom resolution22. Further, in a meta-analysis pooling data from four studies reporting response to antibiotics for MO in CeD, 22/23 (96%) had significant improvement in symptoms and moreover, in the two studies which included breath testing before and after antibiotics, 100% of breath tests normalized after antibiotics5. Finally, a previous study at our institution found that 67% of nonresponsive CeD patients with SIBO had a co-existing disorder (refractory CeD, microscopic colitis, or enteropathy-associated T-cell lymphoma)7, so overgrowth in CeD may represent an epiphenomenon reflective of complicated CeD or an associated disorder. Future studies should prospectively evaluate the risk of MO at the time of diagnosis and throughout the disease course, determine the optimal diagnostic modaility, take into account potential confounders and predictors, and assess response to treatment.

In conclusion, gastrointestinal microbial overgrowth is common among patients with celiac disease, occurring in 17.6–49.6% of cases when diagnosed with intestinal aspirate culture and up to 23.1% with breath test. SIBO is especially common in refractory disease, occurring in 36.8–78.9% of patients. SIFO on the other hand is rare, and if present, most often co-exists with SIBO. There is no association between Marsh score and SIBO positivity, and SIBO does not appear to contribute to villous damage in treated CeD as this was seen commonly in patients with Marsh scores <2. All positive breath tests reflected intestinal methanogen overgrowth. MO should be considered as a cause for persistent symptoms in patients with CeD.

Methods

Patient identification

Utilizing the electronic health record, we identified patients who had undergone MO testing either with intestinal aspirate culture or glucose or lactulose breath testing between October 2021 and September 2023 and who had a diagnosis of CeD. Patients were screened for CeD via ICD codes, and then each patient’s medical record was individually appriased to confirm that the diagnosis of CeD was correct, determined by serologic and histopathologic criteria. Patients without CeD or with other causes for enteropathy were excluded.

MO test protocols

Breath testing protocol involves having patients ingest 10 g of lactulose or 75 g of glucose in solution, and having breath hydrogen and methane measured at baseline every 15 min until 90 min (or 60 min if positive test criteria are already met). For patients who underwent intestinal aspirate culture, Marsh score from the same endoscopy was ascertained. Our institution’s method to obtain aspirates cultures has been described previously7. Briefly, during esophagogastroduodenoscopy, the endoscope is advanced to at least the second part of the duodenum while avoiding suctioning. A sterile 240 cm long by 2.3 mm wide single-lumen catheter (Hobbs Medical, Inc, CT) is introduced and advanced as far distally into the small bowel as possible, at which point the assistant aspirates enteric fluid. The obtained fluid is grown on aerobic trypticase soy agar with blood, eosin methylene blue, anaerobic CDC blood agar, phenylehthyl alcohol agar, and Laked blood with Kanamycin and Vancomycin agar. Results are reported as CFU/mL (none, <1000, 1000–10,000, 10,000–100,000, and >100,000) and breakdown by aerobic versus anaerobic. Microbial speciation is reported when there is a predominance or >100,000 CFU/mL of one or more organisms. Aspirate cultures were categorized as positive by two total bacterial load thresholds: ≥103 colony forming units (CFU) per milliliter (mL) and ≥105 CFU/mL, given ongoing debate regarding the optimal diagnostic threshold14. SIFO was diagnosed in the presence of ≥103 CFU/mL of fungi or yeast42. Breath tests were considered positive if there was a rise in breath hydrogen of ≥20 parts per million (ppm) from baseline by 90 min (or 60 if criteria already met by then) or a breath methane level of ≥10 at any time point in accordance with the North American Consensus20.

Data analysis

Categorical data were summarized with count (percentage) and compared across groups with Chi-Square tests while continuous data were summarized with median (quartile 1, quartile 3). All analyses were conducted in R statistical software (version 4.4.1 Vienna, Austria). This study was approved by the Mayo Clinic Institutional Review Board (study ID 24-007925).

Ethics approval

Approved by Mayo Clinic Institutional Review Board (study ID 24-007925).