Introduction

Irritable bowel syndrome (IBS) is a chronic gastrointestinal disorder with characteristic symptoms of abdominal pain, altered bowel function and bloating. In the Greek population a recent survey among 2,397 individuals revealed that 15.1% of the Greek population suffers from IBS [1]. While the pathophysiology of IBS remains unknown, there is a growing body of research suggesting that gut flora and specifically small intestinal bacterial overgrowth (SIBO) may be implicated in the symptoms of this condition.

Small intestinal bacterial overgrowth (SIBO) is traditionally characterized by excessive bacteria in the small intestine. In addition to quantitative changes in small bowel flora, SIBO is also associated with qualitative changes of intestinal flora. Specifically, bacteria implicated in SIBO are mainly of the colonic type and predominantly Gram-negative aerobe and anaerobe species that ferment carbohydrates into gas. Various conditions predispose to SIBO. These include abnormalities of bowel motility, derangement of bowel innervation, prior gastric surgery and inflammatory bowel diseases [2]. Recently, overproduction of gas in SIBO has also been used to explain the mechanism by which SIBO may lead to symptoms of irritable bowel syndrome like diarrhea and bloating [3]. This hypothesis is fundamental in the attempt to understand a disease with almost pandemic characteristics.

To date, many studies demonstrate an increased prevalence of SIBO among patients with IBS [4–6]. However in all these studies, SIBO is diagnosed by a positive early glucose or lactulose breath test. Since this is an indirect technique for determining SIBO, there has been some controversy as to whether these findings represent transit or SIBO [7]. Thus quantitative cultures of the proximal small intestine are needed to validate this hypothesis since increase of small bowel bacteria is the hallmark of a SIBO diagnosis. Furthermore, breath tests do not provide adequate information on the type of bacteria implicated in SIBO [6].

It should, however, be underscored that the definition of SIBO differs within published articles in relation with the amount of bacterial counts in the small intestine. The purpose of this study was to underscore the frequency of SIBO in a large number of Greeks necessitating upper GI tract endoscopy by using quantitative microbiological assessment of the duodenal aspirate. The study also aimed to further investigate the relationship between SIBO and IBS in the context of co-morbid conditions and drug intake (specifically proton pump inhibitor use).

Patients and Methods

Study Design

The study was conducted from September 2009 until January 2011 in the Department of Gastroenterology of Sismanogleion General Hospital of Athens. The study protocol was approved by the Ethics Committee of the hospital. Every patient was enrolled once in the study after written informed consent.

Inclusion criteria were: (a) age ≥18 years, and (b) clinical indication for outpatient upper GI tract endoscopy. Exclusion criteria were: (a) any antibiotic intake the last month, (b) infections by the human immunodeficiency virus-1 (HIV-1), by the Hepatitis B virus (HBV) or by the hepatitis C virus (HCV), (c) Child Pugh liver cirrhosis stages 2 and 3, (d) active GI tract bleeding, (e) gastroesophageal reflux disease (GORD), (f) presence of systemic sclerosis, and (g) history of hypothyroidism. Patients infected by HIV-1 were excluded because this is a state of immunosuppression that may lead to erroneous results regarding intestinal bacterial overgrowth and local immune system interactions. Patients infected by HBV and HCV were excluded because these infections may lead to liver insufficiency and affect gastric pH. GORD was an exclusion criterion because it may alter gastric pH and possibly affect bacterial growth in the upper small intestine.

The following information was recorded for each patient: age, gender, height, weight, reason for endoscopy, endoscopic findings, other diseases and intake of any medication. Body mass index (BMI) was calculated in all patients by the standard formula [8]. Before undergoing upper GI tract endoscopy, patients were asked to provide replies to a specific questionnaire trying to disclose the presence of IBS. The questionnaire involved enquiry regarding symptoms of abdominal pain or discomfort and the relationship with bowel habits. According to the provided replies, patients were classified as IBS sufferers or not. IBS was defined as recurrent abdominal pain or discomfort at least 3 days per month in the previous 3 months associated with two or more of the following according to the Rome II criteria: (a) improvement with defecation, (b) onset associated with a change in frequency of stool, and (c) onset associated with a change in form (appearance) of stool [9]. Patients with IBS were further classified into three bowel habit subtypes according to Rome II criteria [10]: diarrhea-predominant (IBS-D), constipation-predominant (IBS-C), alternating bowel habit (IBS-A). Classification of IBS by the Rome II criteria was used because these were the most widely applied criteria when the study protocol was submitted for approval by the Ethics Committee.

All enrolled patients were examined by upper GI endoscopy with the use of the same endoscope EG-250WR5 (Fujinon, Willich, Germany) after overnight fasting. The duodenum was entered and fluid was aspirated immediately after the entrance of the endoscope into the third part of the duodenum with the use of washing probes (F8MTK2312230 230 cm × 2.30 cm and F8MSTK1812120 120 cm × 1.80 cm). The upper part of the probe was connected with a sterile syringe of 60 ml for fluid aspiration. The sample was then transported and cultured within the first hour after collection.

Laboratory Investigation

Aspirated duodenal fluid was quantitatively cultured under aerobic conditions. More precisely, one 0.1-ml aliquot of fluid was directly plated onto MacConkey agar (Becton–Dickinson, Cockeysville, MD, USA). Then another 0.1-ml aliquot of fluid was diluted 1:10 four consecutive times into Mueller–Hinton broth and another 0.1 ml of each dilution was plated onto MacConkey agar. The plates were cultured for 24 h at 37°C under aerobic conditions. The number of every isolate was counted in each dilution and multiplied with the appropriate dilution factor. The mean of each multiplication was measured and applied. The number of bacteria was expressed as colony forming units (CFU)/ml by their log10 value. The lower limit of detection was 10 CFU/ml. The study was focused on SIBO by aerobes because it was considered that the collection procedure of fluid from the third part of the duodenum does not ascertain anaerobe conditions.

Identification of Gram-negative species was done by API20E and API20NE systems (BioMérieux, Paris, France). Identification of Staphylococcus spp was done by positive catalase test; isolates positive for coagulase were identified as S. aureus. Identification of Enterococcus spp was done by a negative catalase test and by bile and salt solubility.

Since there is not an internationally accepted definition for SIBO, we defined the presence of more than 103 CFU/ml of colonic type aerobe bacteria in the duodenal aspirate as indicative of SIBO. This definition was based on (a) former publications about the definition of SIBO [11], and (b) on a preliminary cohort of 15 patients undergoing upper GI tract endoscopy for mild dyspepsia. These patients had negative histories for co-morbidities and for drug intake and they did not have any finding on endoscopy. As a consequence they could be considered healthy controls. Quantitative culture of their duodenal aspirates collected as described above yielded ≤10 CFU/ml. Based on the latter finding it was considered that isolation of any species at 100-fold counts greater than those of healthy controls may be interpreted as diagnostic of SIBO. However, the above diagnostic cut-off for SIBO is not universally accepted. To this end, some of the results are given using diagnostic cut-offs of ≥104 and ≥105 CFU/ml.

Study Endpoints and Sample Size Calculation

The primary endpoint of the study was to assess the frequency of SIBO by aerobe bacteria in a cohort of patients undergoing upper GI tract endoscopy. Factors related with SIBO were secondary endpoints. The number of patients needed to this end was based on former results that 15.1% of the Greek population reported symptoms compatible with IBS [1]. If a relationship exists between SIBO and IBS, it is calculated that 300 patients should be enrolled with 80% power to identify 50 patients with SIBO.

Analysis of Data

Patients were divided as those with SIBO and those without SIBO by aerobic bacterial species. Isolation of coagulase-negative Staphylococcus spp, viridans Streptococcus spp or any other species indicative of oral bacterial flora or of upper airways bacterial flora was excluded from analysis since it was considered non-diagnostic of SIBO.

Data of age and BMI were expressed by their mean ± standard deviation (SD); those of bacterial growth by their mean ± SE. Normality of distribution was assessed after Kolmogorov–Smirnov’s statistics. Comparisons were done by the Student’s t test for variables with normal distribution of values and by the Mann–Whitney U test for variables with non-normal distribution of values. Comparisons between qualitative data were done by the χ2 test. Odds ratios (OR) and 95% confidence intervals (CI) were calculated by Mantel–Haenszel’s statistics. Step-wise forward logistic regression analysis was performed with presence of SIBO as a dependent variable and diagnosis of IBS, presence of the co-morbidities, consumption of drugs affecting pH and endoscopic diagnosis of gastritis as independent variables. ORs and 95% CIs were calculated. In order to define predictors of SIBO among IBS sufferers, univariate analysis was performed within patients with IBS; SIBO was a dependent variable and a large array of clinical symptoms and co-morbid conditions, history of intake of medications and endoscopic findings as independent variables. The true impact of any predictor for which significance was disclosed by univariate analysis was confirmed by step-wise logistic regression analysis. Any value of P below 0.05 was considered significant. Analysis was done by the PASW Statistics 18.0 software (SPSS for Windows version 18.0).

Results

Patients’ Demographics

A total of 518 patients were screened for eligibility; 320 patients were enrolled in the study (Fig. 1). SIBO by aerobe species was diagnosed in 62 patients (19.4%). Demographic characteristic of all patients enrolled in relation with the presence or absence of SIBO are shown in Table 1. Demographically, subjects with SIBO were younger, more often had a history of diabetes mellitus type 2 (DM2) and were more often taking proton pump inhibitors (PPIs) and acenocoumarone. Although not reaching statistical significance, there was also a trend for subjects with SIBO to have a greater BMI.

Fig. 1
figure 1

Study flow chart

Full size image
Table 1 Demographic characteristics of patients enrolled in the study
Full size table

The most common symptoms of patients with SIBO were bloating in 44 patients (70.9%), diarrhea in 21 patients (33.9%), and flatulence in 21 patients (33.9%). None of these patients had history of steatorrhea or of a recent CI tract infection.

Using a diagnostic threshold of more than 104 CFU/ml of colonic type aerobe bacteria in the duodenal aspirate, SIBO was diagnosed in 52 patients (16.3%). Using a diagnostic threshold of more than 105 CFU/ml of colonic type aerobe bacteria in the duodenal aspirate, SIBO was diagnosed in 35 patients (10.9%).

SIBO and IBS

Of the initial 320 subjects, 112 were IBS sufferers; 35 patients (31.2%) had IBS-D; 19 patients (16.9%) had IBS-C; and 58 patients (51.8%) had IBS-A. Reasons for upper GI tract endoscopy among IBS patients were dyspepsia in 75 patients (66.9%), anemia in 24 patients (21.4%) and change of bowel frequency in nine patients (8.0%); the remaining patients had a list of other uncommon complaints. Fifty-eight of these patients (51.8%) had co-morbidities; the most common co-morbidity was diabetes mellitus type 2 in 13 patients (11.6%) and previous solid tumor malignancies in 11 patients (9.8%); other co-morbidities were seen in less than 10% of subjects.

Among these IBS subjects, 42 (37.5%) met the applied definition for SIBO. Examining the relationship between SIBO and IBS, among the 62 patients with SIBO, 42 had IBS (67.7%). This was greater than the 258 patients without SIBO, whereby only 70 had IBS (27.1%) (Mantel–Haenszel common odds ratio estimate: 5.64; 95% CI 3.09–10.27; P < 0.0001).

When using the 104 CFU/ml diagnostic threshold for SIBO, it was found that IBS was present in 35 out of 52 patients with SIBO (67.3%) and in 77 out of 268 patients without SIBO (28.7%) (P of comparison <0.0001). When using the 105 CFU/ml diagnostic threshold for SIBO, it was found that IBS was present in 24 out of 35 patients with SIBO (68.6%) and in 88 out of 285 patients without SIBO (30.9%) (P of comparison <0.0001).

The frequency of SIBO was much greater among patients with IBS-D. More precisely, 33.8% of patients with SIBO were sufferers of IBS-D compared with 5.4% of patients without SIBO (P < 0.0001) (Table 1).

Based on results presented in Table 1, it was decided that history of DM2 should be included in logistic regression analysis among all co-morbid conditions; in the same sense consumption of PPIs should be included in analysis among all types of drugs affecting gastric pH. Logistic regression analysis revealed that all factors entering the equation had statistical significance (Table 2). After controlling for DM2 and PPI use, IBS was still the most important factor associated with SIBO. More precisely, IBS and consumption of PPIs were favoring SIBO. On the contrary, the presence of gastritis was associated with absence of SIBO.

Table 2 Step-wise forward logistic regression analysis of factors linked with small intestinal bacterial overgrowth (SIBO) by aerobe bacteria
Full size table

Predictors of SIBO in IBS

Univariate analysis revealed (Table 3) that age, history of chronic heart failure, history of chronic renal failure, presence of gastritis, consumption of PPIs, consumption of non-steroidal anti-inflammatory drugs, consumption of low dose aspirin daily, consumption of acenocoumarone and presence of diarrhea were related with SIBO.

Table 3 Univariate analysis of a large array of clinical symptoms, of co-morbid conditions, of drug consumption and of endoscopic findings related with the presence of syndrome of intestinal bacterial overgrowth in 112 patients with irritable bowel syndrome
Full size table

To fully define the impact of these factors of the advent of SIBO within patients with IBS, step-wise logistic regression analysis was done (Table 4). Results revealed that the only factors associated with SIBO in IBS were age, history of chronic renal failure and the presence of diarrhea. The presence of chronic renal failure was preventive against SIBO whereas age and diarrhea were positively linked with SIBO.

Table 4 Step-wise forward logistic regression analysis of factors linked with small intestinal bacterial overgrowth (SIBO) by aerobe bacteria in 112 patients with irritable bowel syndrome
Full size table

Types of Bacteria and SIBO

The most common bacterial species isolated from the duodenal aspirate of the 62 patients with SIBO were Escherichia coli (n = 23; 37.1%), Enterococcus spp (n = 20; 32.3%), Klebsiella pneumoniae (n = 15; 24.2%), Proteus mirabilis (n = 4; 6.5%), Acinetobacter baumannii (n = 3; 4.8%), Citrobacter freundii (n = 3; 4.8%), Serratia marscecens (n = 3; 4.8%), Staphylococcus aureus (n = 2; 2.9%), Pseudomonas putida (n = 2; 2.9%), Pasteurella multocida (n = 2; 2.9%) and Enterobacter aerogenes (n = 1; 1.6%).

Fifteen patients with SIBO were carriers of more than one bacterial species. More precisely, six patients were carrying E. coli and K. pneumoniae, three patients were carrying E. coli and Enterococcus spp, three patients were carrying E. coli and P. mirabilis, one patient was carrying Enterobacter cloacae and Enterococcus spp, one patient was carrying P. mirabilis and Acinetobacter baumannii, and another patient was carrying E. cloacae and Citrobacter freundii.

Total bacterial growth in the duodenal aspirates of patients with IBS was greater than among patients without IBS (P < 0.0001, Fig. 2). That was also the case for patients with SIBO compared with patients without SIBO (P < 0.0001). However, total bacterial growth in the duodenal aspirates was similar between patients with SIBO and IBS and those with SIBO without IBS.

Fig. 2
figure 2

Total bacterial growth in the duodenal aspirates of patients enrolled in the study. Patients are divided in relation with the presence or not of syndrome of intestinal bacterial overgrowth (SIBO), in relation with the presence or not of irritable bowel syndrome (IBS) and in relation with the presence of SIBO with IBS and SIBO without IBS. Mean values and standard errors (SE) are indicated. P values refer to indicated comparisons

Full size image

Discussion

In this study, the only gold standard technique known for the detection of small intestinal bacterial overgrowth (small bowel culture) is used to examine the prevalence of SIBO in IBS compared to subjects with other GI disorders undergoing endoscopy. SIBO was seen more often in IBS and up to 60% of IBS subjects with IBS-D. A strong association of SIBO with IBS was found even when the diagnostic cut-off for SIBO was increased to ≥105 CFU of bacteria/ml. In addition, the overall bacterial counts were higher in IBS compared to all other subjects. After controlling for presence of DM2 and intake of PPI, IBS was still associated with SIBO based on culture.

The contribution of SIBO in the pathogenesis of IBS had led to several trials whereby administration of non-absorbable antibiotics was tested and demonstrated improvement of symptoms of IBS [12–15]. The most successful of all these antibiotics is rifaximin which has two major characteristics: (a) limited systemic absorption and (b) antimicrobial spectrum comprising both aerobic and anaerobic species implicated in SIBO. In one double-blind, randomized, placebo-controlled phase II study of 87 patients, oral administration of rifaximin 400 mg tid for 10 days was accompanied by overall improvement of symptoms of IBS; the benefit persisted for 10 weeks after the end of treatment [16]. Two randomized clinical trials have recently been published, TARGET 1 and TARGET 2, where oral 550 mg rifaximin tid was administered. Both trials had similar design and enrolled patients without constipation; 623 patients were assigned to TARGET 1 and 637 patients to TARGET 2. Global IBS symptoms and abdominal pain and bloating were significantly improved [17].

The efficacy of non-absorbable antibiotics supports the hypothesis of altered gut flora in IBS as a major contributor in the pathogenesis of IBS but only indirectly supports SIBO specifically. The specific description of SIBO in IBS has been based on several prospective observational studies of patients suffering from IBS with the application of either lactulose or of glucose breath tests [4–6]. According to these studies, the prevalence of SIBO is noted to be as high as 84% among IBS sufferers but depends on the study and type of substrate used for breath testing. Lactulose studies tend to have higher rates of SIBO on breath testing. A recent meta-analysis of 12 studies enrolling 1,921 patients with IBS diagnosed either with Rome I or Rome II criteria showed 54% pooled prevalence of positive lactulose breath test and 31% pooled prevalence of positive glucose breath test. However, pooled prevalence of positive jejunal cultures was only 4% [18]. The results of our study demonstrate that overall, 37.5% of IBS subjects have overgrowth from aerobe colonic type bacteria in the small intestine. That is even greater at 60% among patients with IBS-D.

The above data lead to question which signs may predict for SIBO among patients with IBS. In a study of 98 patients with IBS of which 36% had SIBO as defined by a positive glucose breath test, female gender and age were considered predictors of SIBO [19]. In another survey of 59 patients with IBS, female gender, bloating and diarrhea were strong predictors of SIBO [20]. In the present study, age and diarrhea were the strongest predictors of SIBO within the population of IBS patients. The analysis presented here for diarrhea as one predictive symptom for SIBO within patients with IBS reinforces the importance of the beneficial results of rifaximin administration in patients with IBS without constipation [17].

While the gold standard for the diagnosis of SIBO is quantitative culture of bacteria from the proximal small intestine, this can be difficult for many reasons: (a) the risk of contamination by bacterial flora of the oral cavity and of the upper respiratory tract, (b) anaerobe culture of specimens exposed to an aerobic environment prior processing, (c) interpretation of data [21], (d) access to only the most proximal small intestine, and (e) this method is only of research interest for the study of pathophysiology of patients but is too difficult for everyday clinical practice. To overcome these difficulties, isolates compatible with bacterial flora of the upper airways and of the oral cavity were excluded from analysis.

Some main limitations of the present study should be addressed. First, the lack of enrolment of healthy controls. This was difficult for ethical reasons. To this end, a preliminary cohort of patients with mild dyspepsia without any co-morbidities and drug intake who underwent similar sampling was reported. Second, duodenal aspirates were not cultured under anaerobic conditions. However, in the case of only access to the most proximal small intestine, this may only underestimate the true prevalence of SIBO in the IBS population. Third, breath test was not performed in the study population to correlate results with cultures of the duodenal fluid. While breath testing was not conducted in this study, the disclosed rate of 67% for SIBO in IBS is similar to the prevalence rates reported in breath testing studies [4–6]. Fourth, the age range of patients enrolled in the present study creates the impression of an overall elderly population. However, patients of the present study are of a similar age with those enrolled in two recent studies [19, 20]. One of these studies [19] and the present study denote older subjects with IBS to be prone for SIBO. And fifth, thorough laboratory search to exclude the presence of other diseases causing SIBO was not done. Among them duodenal biopsy to exclude celiac disease, stool microscopy to exclude recent GI tract infection, enteroclysis to exclude the presence of intestinal diverticulae and intestinal fistulae and antroduodenal manometry to exclude GI tract dysmotility are the most important. However it should be underscored that none of the patients with SIBO had steatorrhea or recent GI tract infection. Furthermore, the relationship of SIBO with IBS was proved after step-wise logistic regression analysis controlling for the presence of type 2 diabetes mellitus which is a major cause of gut dysmotility.

To our knowledge, only one other study has tried to culture aspirates of the proximal small intestine, more precisely those of the jejunum [22]. A total of 162 patients with IBS were enrolled and SIBO was defined as the presence of more than or equal to 105 of colonic type bacteria/ml. Jejunal content was cultured under both aerobic and anaerobic conditions; anaerobes were rarely isolated allowing to presume that the lack of anaerobe cultures in the present study does not underestimate the prevalence of SIBO. In the same study, SIBO was diagnosed in only 4% of patients with IBS using >105 CFU/ml; however 43% of IBS patients were carriers of more than 5 × 103 bacteria/ml rendering the dilemma that the detection limit for the diagnosis of SIBO should be decreased to 103 colonic type bacteria/ml. Although SIBO was defined as the presence of more than 103 CFU/ml in the duodenal fluid, two other detection thresholds of more than 104 CFU/ml and of more than 105 CFU/ml were studied. Results revealed a positive relationship between SIBO and IBS even when these higher thresholds were applied. The discrepancy between the reported findings and those of Posserud et al. [22] may be explained by a number of methodological issues [22]. First, in that study aspirates were taken after 6 h of water perfused manometry which inevitably diluted the jejunal contents. Second, the authors chose not to use their own controls as the benchmark for SIBO but rather historical data of >105 CFU/ml. It should however be mentioned that quantitative cultures of jejunal aspirates were also reported in 12 patients with IBS who were used as comparators in a study enrolling patients with tropical sprue. Growth of aerobe Gram-negative bacteria was detected in three of them [23].

It is questioned whether SIBO is the real mechanism leading to IBS or an epiphenomenon of other conditions like intestinal dysmotility or intake of PPIs that alter gastric pH and favour bacterial growth [24]. Linkage of SIBO with the pathogenesis of IBS is argued by findings of gut mucosa inflammation of patients with IBS [25]. The shear contribution of SIBO in IBS is proved in the present study by step-wise logistic regression analysis where SIBO remains the major factor positively linked with IBS even in the presence of other conditions like DM2, intake of PPIs and presence of gastritis.

In summary, this study demonstrates a greater prevalence of overgrowth by aerobic bacteria in the small intestine of patients with IBS in this analysis of a prospective cohort of patients undergoing upper GI tract endoscopy. The rate of SIBO is as high as 60% among IBS-D sufferers. Interestingly, SIBO was independently linked with the presence of IBS, history of DM2 and intake of PPIs whereas presence of gastritis was preventive from SIBO. These findings underscore the need for a two-arm prospective study among sufferers and non-sufferers for IBS to disclose the difference in the prevalence of SIBO. This would also allow documentation of the prevalence of SIBO within patients with different IBS sub-classes so as to provide further evidence for the therapeutic role of non-absorbable antibiotics.