Introduction

Because of the increasing prevalence of obesity [1], non-alcoholic fatty liver disease (NAFLD) has become one of the major causes of liver diseases [2]. The spectrum of liver injury is broad, from steatosis and non-alcoholic steatohepatitis (NASH) through to cirrhosis, liver failure and hepatocellular carcinoma [3, 4], and almost all morbidly obese patients undergoing bariatric surgery have some liver lesions [5]. The prevalence of steatosis in these patients is very high, ranging from 85% to 98%, while the prevalence of NASH is highly variable, with an overall prevalence of 37% (24–98%) in a recent review [6]. Diabetes mellitus and insulin resistance are the conditions most frequently associated with NASH and hypertension with advanced hepatic fibrosis [6, 7].

Several data in the 1970s suggested that small intestinal bacterial overgrowth (SIBO) could play a role in NAFLD. Descriptions of fatty degeneration have been reported in morbidly obese patients with jejuno-ileal bypass [8, 9] and in small bowel diverticulosis [10], both conditions favouring bacterial overgrowth. Wigg et al. [11] and Sajjad et al. [12] showed that the prevalence of SIBO assessed by (14)C-D-xylose and lactulose or glucose breath tests was higher in patients with NASH compared with control subjects. However, in patients with morbid obesity, the prevalence of SIBO is unknown and its role in liver injuries has not been studied before. Our aim was, then, to determine the prevalence of SIBO in a prospective cohort of morbidly obese patients referred for bariatric surgery and to study the relationship between bacterial overgrowth and liver histology.

Patients and Methods

Study Design

This prospective study was part of an extensive preoperative and perioperative data collection on patients with morbid obesity referred to our institution to have adjustable gastric bands inserted laparoscopically or to undergo gastric bypass for weight reduction following an unsuccessful hypocaloric regimen. Morbid obesity was defined as body mass index (BMI) ≥  40 kg/m2 or ≥  35 kg/m2 in association with other comorbidities, such as sleep apnoea, diabetes mellitus or cardiovascular disease.

All obese patients systematically had a glucose hydrogen breath test before surgery to look for small intestinal bacterial overgrowth. A liver biopsy, taken surgically in the left lobe as a routine part of the operative procedure, was assessed for specific histological abnormalities. We looked for a relationship between SIBO and the presence of liver lesions.

We also performed a glucose breath test in a group of healthy non-obese (18.5 kg/m2<BMI<25 kg/m2) subjects matched for age and sex with the group of patients to compare the prevalence of positive breath test in the two groups.

Patients and Preoperative Assessment

Preoperative assessment included clinical and familial assessment, psychiatric assessment to detect chronic alcohol consumption and/or compulsive eating disorders, anthropometric measurements (weight, height, waist and hip circumferences) and laboratory tests. Increased waist-to-hip ratio (WHR), reflecting central obesity, was defined as ≥  0.90 in men and ≥  0.85 in women [13]. Laboratory tests included plasma aminotransferases (alanine aminotransferase [ALT], aspartate aminotransferase [AST]) and gamma glutamyltransferase (GGT) activity (upper normal limits of 55, 45 and 50 IU/L, respectively). Other laboratory tests included iron studies, fasting lipid profile, fasting glucose profile, hepatitis B and C serologies, anti-smooth muscle antibodies, anti-liver–kidney microsome antibodies and anti-mitochondrial antibodies. Diagnosis of type 2 diabetes was based on the American Diabetes Association criteria [14]. Dyslipidaemia was defined as an increase in triglycerides or cholesterol or both, or if the patient was taking a hypolipaemic agent. Hypertension was diagnosed if the patient had a past history of hypertension and was on antihypertensive medication or if the patient had a resting recumbent blood pressure greater than or equal to 140/90 mmHg on two repeated occasions. The NCEP-ATPIII definition was used to determine the presence or absence of the metabolic syndrome [15].

Patients were excluded from the study if they had a history of compulsive eating disorders, a daily consumption of alcohol greater than 20 g/day, evidence of hepatitis B or C, a positive serology for anti-smooth muscle antibodies or anti-liver–kidney microsome antibodies or anti-mitochondrial antibodies, any other specific liver disease, a severe associated pathology, previous bariatric surgery or were taking known hepatotoxic medication, such as amiodarone, diltiazem or corticosteroids.

Informed written consent was obtained from every patient and the study was conducted in accordance with the Helsinki Declaration.

Glucose Hydrogen (H2) Breath Test

Subjects fasted overnight (12 h) and during the H2 breath test. All subjects were asked to avoid eating food containing fibre during the previous evening, as this may cause prolonged excretion of H2. Cigarette smoking and exercise were not allowed for at least 2 h before and during the test. End-expiratory breath samples were obtained at 30-min intervals before and for 4 h after an oral load of 50 g of D-glucose, as described previously [16–18]. The results were expressed in parts per million (ppm). A positive result was defined as a fasting breath H2 concentration above 20 ppm and or a breath H2 concentration increase of more than 12 ppm over baseline within the first 2 h of testing [19, 20]. H2 concentrations of expiratory gas samples were measured using an electrochemical cell (Exhaled Hydrogen Monitor; GMI Medical Ltd., Renfrew, Scotland). The minimal detectable H2 concentration was 1 ppm. Equivocal tests were considered as negative. Tests were not interpretable if antibiotics had been taken or a bowel washout was performed one month prior to testing.

Histological Assessment

All liver biopsies were fixed and examined using haematoxyline-eosin, Masson trichrome, silver reticulin and Perl’s stain. Two pathologists, blinded to the patients’ clinical condition and biochemical data, reviewed every biopsy. Steatosis was graded on a 0 to 3 semi-quantitative scale [21]: (0) no steatosis; (1) up to one third of the hepatocytes with steatosis; (2) from one to two thirds of the hepatocytes with steatosis; (3) more than two thirds of the hepatocytes with steatosis, defining severe steatosis. Liver fibrosis was assessed on a five-stage scale [21]: (0) no fibrosis; (1) perisinusoidal fibrosis without portal or periportal fibrosis; (2) perisinusoidal fibrosis with portal or periportal fibrosis; (3) perisinusoidal fibrosis with portal or periportal fibrosis with focal or extensive bridging fibrosis; (4) cirrhosis. Lobular necroinflammatory activity was defined as the presence of lobular polymorphonuclear leukocytes infiltrate associated with lobular necrosis. Non-specific inflammation was defined as sparse to mild focal inflammation without necrosis. The presence of ballooning degeneration and Mallory bodies was described.

Statistical Analysis

The results are expressed as mean±standard error (SE) and percentages. Comparisons of quantitative variables were made using Student’s t-test. Categorical variables and proportions were tested by the χ2 and Fisher’s exact tests. The relationship between severe steatosis or NASH and various risk factors was studied using univariate comparison followed by multivariate analysis using binary logistic regression. Baseline variables that reached a univariate P-value ≤  0.15 and variables found to be predictive in the literature were included in the multivariate analysis. A multivariate analysis was performed using either severe steatosis or NASH as the dependent variables. A P-value <  0.05 was considered as statistically significant. Analysis was performed with the SPSS software package (version 11.5; SPSS Inc., Chicago, IL).

Results

Characteristics of Patients

From October 2001 to October 2004, 146 consecutive patients with morbid obesity (129 F/17 M) who had a glucose H2 breath test were included in the study. The mean age was 40.7±11.4 years (range: 17–72 years) and the mean BMI was 46.1 ±  6.4 kg/m2 (range: 35.4–73.3 kg/m2). Hypertension was present in 29.2%, diabetes mellitus in 19.8% and dyslipidaemia in 17.7% of the patients. A metabolic syndrome was present in 46.3% of the patients. The mean waist size was 120.3±10.7 cm for females and 138.1±11.9 cm for males. The mean WHR was 0.86±0.09 for females and 1.01±0.05 for males. As liver tests were not routinely performed initially, pre-operative values for aminotransferases were only available for 80 patients. The mean liver enzyme levels were 44±17 IU/L for ALT, 23±15 IU/L for AST and 48 ±  35 IU/L for GGT.

Glucose Hydrogen Breath Test

A glucose hydrogen breath test was performed in 40 healthy volunteers, which was positive in one case (2.5%). Of the initial 146 breath tests performed in obese patients, six were excluded because of antibiotic intake within the previous month. Of the remaining 140 breath tests, 24 were positive (17.1%), 112 were negative (80%) and four (2.9%) were equivocal because the increase of H2 was not sufficient or not sustained. The frequency of positive breath tests was significantly higher in patients than in healthy subjects (P=0.031). The characteristics of the patients according to positive or negative breath tests are presented in Table 1. In the univariate analysis, no variable was associated with the presence of SIBO.

Table 1 Characteristics of the patients according to positive or negative breath tests
Full size table

Histopathological Findings

Surgery was not performed in nine patients because of various contra-indications. A liver biopsy was available in 137 out of the 146 patients of the initial group.

Steatosis was present in 126 patients (92%) (grade 1 in 47.2%, grade 2 in 29.6% and grade 3 in 15.2%). According to the five-stage scale score of fibrosis, fibrosis was present in 72 patients (52.6%) (grade 1 in 28.5%, grade 2 in 13.1% and grade 3 in 0.7%). Isolated portal fibrosis was present in 14 patients (10.2%), lobular necroinflammatory activity in 54 patients (39.4%) and NASH in 34 patients (24.8%). Five patients (3.7%) had normal histology.

The histopathological findings according to positive or negative breath tests are presented in Table 2. When SIBO was present, there was a small increase (from 7 to 16%) in the frequency of sinusoidal fibrosis, lobular necroinflammatory activity and NASH, but these increases never reached a level of statistical significance. The frequency of severe steatosis tended to be increased in patients with SIBO (P=0.127). Other factors associated with severe steatosis in the univariate analysis are presented in Table 3. Severe steatosis was associated with age, serum triglyceride level and the presence of a metabolic syndrome, with also a trend for higher ALT level (P=0.066). The multivariate analysis (binary logistic regression) for the presence of severe steatosis is presented in Table 4. The presence of a metabolic syndrome and SIBO were independent predictive factors for severe steatosis (the predictive value of ALT in the multivariate analysis could not be tested because of too much missing data).

Table 2 Histological findings according to positive or negative breath tests (n=127)
Full size table
Table 3 Univariate analysis of the predictive factors for severe steatosis (>66% of hepatocytes)
Full size table
Table 4 Multivariate analysis (binary logistic regression) for the presence of severe steatosis
Full size table

In the univariate analysis, the presence of NASH was associated with higher waist circumference (127.5±15.4 cm vs. 120.7±11.0 cm, P  =  0.037) but not with age (41.6±11.4 years vs 39.7±11.4 years, P  =   0.40), sex (82.4% vs. 90.3% of women, P   =   0.35), BMI (47.2±6.7 kg/m2 vs. 46.1±6.4 kg/m2, P   =   0.4) or the presence of a metabolic syndrome (52.6% vs 41.9%, P=0.4). In the multivariate analysis (binary logistic regression), only waist circumference (OR 1.06. CI [1.001–1.116], P=0.047) was found to be predictive for the presence of NASH, whereas age, sex, BMI, the presence of metabolic syndrome and SIBO were not predictive.

Discussion

In the present study, we found that bacterial overgrowth, defined by a positive glucose H2 breath test, was more frequent in morbidly obese patients compared to healthy subjects and was associated with severe steatosis.

The commonly accepted definition of SIBO is a total growth of >105 colony-forming units/ml of intestinal fluid. Because of the difficulties in obtaining cultures, different breath tests, including the glucose H2 breath test, have been used. In our study, 17% of patients with morbid obesity had a positive glucose breath test; this was significantly higher than in healthy subjects. In the literature, the prevalence of the positive glucose breath test varies depending on the definition used for a positive breath test [22]. The prevalence of 2.5% of positive breath test found in our healthy subjects is similar to that reported before, varying from zero to less than 5% [20, 23–25]. We did not find any clinical or biochemical patient characteristic associated with SIBO. In the literature, older age has sometimes been associated with the increased prevalence of SIBO [25–27], but in our study, only four patients were greater than 60 years of age. Another known risk factor of SIBO is the use of proton pump inhibitors [27]. Because of elevated gastroesophageal reflux disease (GERD) prevalence in morbidly obese patients [28, 29], it could have been a confounding factor in our series, but proton pump inhibitors were stopped 7 days before the breath tests were performed. Bacterial overgrowth in morbidly obese patients could also be due to intestinal motility alterations, as described previously in humans or mice [30, 31]. Intestinal motility was not assessed in our patients, but one study performed in patients with severe obesity found motility disturbance in the fasting period with diminished Phase I, increased Phase II and more distal and less frequent occurrence of Phase III activity of the migrating motor complex [32].

A quarter of our patients had NASH and 92% had steatosis, which was severe in 15.2% of them. These findings are in accordance with the literature [5, 6]. In the presence of SIBO, the prevalence of portal fibrosis, sinusoidal fibrosis, lobular necroinflammatory activity and NASH was similar or slightly higher, but the difference never reached a level of statistical significance. However, there was a trend for an increased prevalence of severe steatosis from 10 to 26% of cases in the presence of SIBO, and SIBO was also found to be an independent predictive factor of severe steatosis. In the univariate analysis, other factors associated with severe steatosis were older age, serum triglyceride level and the presence of metabolic syndrome, but in the multivariate analysis, apart from SIBO, only the presence of a metabolic syndrome was found to be an independent predictive factor. Concerns about the variability of liver biopsy in NAFLD have been previously raised. However, in morbidly obese patients, excellent agreement for steatosis was seen in biopsy specimens obtained at surgery from the right and left lobes of the liver [33, 34]. Recently, Mathurin et al. [35] found that, before bariatric surgery, severe steatosis was associated in the univariate analysis with ALT, GGT, fasting blood glucose, serum triglyceride levels and insulin resistance index, while in the multivariate analysis, only ALT level and insulin resistance index were independent predictive factors [35]. At the time of our study, insulin resistance testing was not routinely performed, but, despite no strict overlapping, a metabolic syndrome is significantly associated with insulin resistance [36].

The link between SIBO and steatosis is supported by a complete or partial reversal of steatosis after metronidazole treatment in patients with intestinal bypass [37] and in various rat models of SIBO [38, 39]. The mechanisms of liver lesions produced by bacterial overgrowth have been studied in rodent models. The peptidoglycan-polysaccharide, a bacterial cell wall polymer with potent inflammatory and immunoregulatory properties, seems to have a crucial role in promoting liver injuries. In a rat model of SIBO, treatment with mutanolysin, a highly specific muralytic enzyme, prevented the elevation of both plasma anti-peptidoglycan antibodies and tumour necrosis factor-alpha (TNF alpha) levels in Kupffer cells stimulated in vitro with peptidoglycan-polysaccharide [40]. The recent data of Cani et al. [41, 42] in mice also suggest that gut microbiota contribute to the pathophysiological regulation of endotoxaemia and may dysregulate the inflammatory tone, leading to insulinoresistance and obesity. Moreover, in models of genetically obese rats and mice, there is an increased sensitivity to endotoxin hepatotoxicity caused by hepatic macrophage dysfunction [43]. In ob/ob mice, a model of NAFLD that develops intestinal bacterial overgrowth and overexpresses TNF alpha, treatment with probiotics or anti-TNF antibodies improves liver histology and reduces hepatic total fatty acid content [44] by reducing a TNF alpha kinase and the DNA binding of nuclear factor kappa B that promoted insulinoresistance. These data and our results suggest that SIBO could play a role in the initial phase of steatohepatitis in the “two-hit” theory proposed by Day and James [45]. Severe steatosis constituting the “first hit,” the progression to more severe lesions, such as NASH, could be related to a “second hit” caused by oxidative stress/lipid peroxydation [46] and by factors promoting the expression of proinflammatory cytokines [47].

Conclusion

We found, in a population of morbidly obese patients, that bacterial overgrowth was more frequent than in healthy non-obese subjects and was associated with severe hepatic steatosis. Further studies are necessary to confirm our findings and to determine the mechanisms of such bacterial overgrowth. If our data are confirmed, this could open up a new field of treatment with probiotics in order to limit the progression towards more severe liver lesions, such as non-alcoholic steatohepatitis (NASH) and cirrhosis.