During November 2010 and March 2012, 46 consecutive hospitalized patients with liver cirrhosis, defined by the combination of clinical, analytic and imaging criteria, were included in the study. Exclusion criteria were gastrointestinal bleeding, the use of antibiotics, steroids or immunosuppressive drugs or drinking alcohol within the last two weeks. Written informed consent was given by all patients. The study protocol was approved by the Internal Review Board (Ethics Committee of the Friedrich Schiller University Jena, 2931-09/10). Etiology of the liver cirrhosis, age, gender, grade of ascites and hepatic encephalopathy, concurrent medication, history of SBP and vital parameters of the patients were assessed on the day of inclusion. Components of the model for end-stage liver disease (MELD) score and the Child-Pugh score were assessed using routine laboratory data. To assess intestinal permeability, lactulose/mannitol test was performed as described below. Serum samples were collected at the day of inclusion and stored at -80 °C until further analysis. Patients were followed-up for a maximum of 365 d and assessed for survival, liver transplantation, the necessity of hospitalization and the occurrence of bacterial infections. As a control group, 16 healthy volunteers were enclosed. All of them had an inconspicuous medical history without gastrointestinal abnormalities, infections, liver disease, history of major intestinal surgery or alcohol abuse and were not allowed to have taken antibiotics or anti-inflammatory drugs within the last month.

The primary end-point of the study was defined as differences in 90-d mortality rates between patients with pathological and without pathological lactulose/mannitol test. Owing to few events within 90 d (4 deaths), the observation period was extended to one year and differences in survival between both groups were analyzed. Secondary endpoints were the occurrence of bacterial infections, occurrence of SBP and the number of hospitalizations.

After overnight fasting, the patients and controls were instructed to drink 100 mL of a solution consisting of 2 g mannitol and 5 g lactulose with 40 g glucose as an osmotic filler[21,22]. Concomitant medication with lactulose was stopped two days before the test. Urine was collected over 5 h and the volume was measured. Two aliquots of 9 mL urine each were centrifuged for 5 min (700 rcf, 20 °C) and cell-free supernatants were stored at -20 °C. High performance liquid chromatography analysis (Hitachi High Technology, Schaumburg, IL, United States) was performed by GANZIMMUN Diagnostics AG^® (Frankfurt, Germany) to determine the lactulose and mannitol concentration in the 5 h urine collection. The IP index (lactulose/mannitol ratio) was calculated for each study subject using the following formula:

IP index = recovery_lac/recovery_man =

c_lac× V/5 g/c_man× V/2g

Pathologically increased IP was defined by an IP index greater than 0.07[23].

From each patient, a maximum of 9 mL of whole blood were taken, allowed to clot at room temperature over 30 min, centrifuged for 10 min at 1000 rcf and serum was stored at -80 °C until further analysis. Enzyme-linked immunoassays (ELISA) were used to asses the concentrations of intestinal fatty acid-binding protein (I-FABP), lipopolysaccharide-binding protein (LBP) and interleukin-6 (IL-6) in serum. I-FABP concentrations were determined in diluted samples using the human I-FABP ELISA Kit (Hycult biotech, Uden, Netherlands) providing a lower limit of detection (LOD) of 47 pg/mL. IL-6 was determined using the human IL-6 ELISA kit (RayBiotec, Norcross, United States) in undiluted samples (LOD 3 pg/mL). LBP was analyzed using the human LBP ELISA Kit (Hycult biotec, Uden, Netherlands) using serially diluted samples (LOD 4.4 ng/mL) as described[24]. Absorbance was measured in 96-well plates on a photospectrometer (Infinite^® 200 Pro, TECAN, Männedorf, Switzerland) at 450 nm. Sample volumes were insufficient to determine I-FABP concentration in 10 samples.

The statistical analyzes were performed using SPSS 22 (IBM, Armonk, NY, United States) and Prism 5 (GraphPad, La Jolla, CA, United States). Bivariate nonparametric correlation analysis (Spearman’s rho, r_s) was performed to identify correlations between IP index and continuous or ordinal variables. To compare two groups we used Wilcoxon-Mann-Whitney tests for continuous and Fisher’s exact test for nominal variables. Time-to-event variables were displayed by the Kaplan Meier method and groups were contrasted by Log-rank tests and Cox regression analyses. Continuous variables were assessed as linear and dichotomized predictors according to the maximum Youden index in receiver operating characteristics (ROC) analysis or presented as quantiles if not assessable in this way. Multiple stepwise-forward Cox regression analysis was performed using dichotomized variables, which met P < 0.05 in univariate analysis and which were available for the total cohort. We applied a two-sided significance level P = 0.05 and did not correct for multiple testing due to the limited statistical power of this explorative observational study.

Forty-six patients with liver cirrhosis were included. Sixty-one percent (28/46) were male. The median age was 59 years (range, 42-81) and the majority of patients had alcoholic liver disease (72%, 33/46) with advanced liver disease. The median Child-Pugh score was 9 (range, 5-14), the median MELD score 15 (range, 6-27). Frequent comorbidities were diabetes (22%, 10/46) and hepatocellular carcinoma (20%, 9/46) (Table 1). Among the 16 controls included, 44% (7/16) were males. The median age of controls was 33 years (range, 26-59).

Thirty-nine (85%) patients with cirrhosis had pathologically increased IP in contrast to four (25%) of the healthy controls (P < 0.0001). In comparison to controls patients with cirrhosis had higher lactulose recovery (median 2.0% vs 0.6%, P = 0.005) alongside lower mannitol recovery (11.2% vs 16.1%, P = 0.03). The IP index was highest in patients with Child-Pugh C cirrhosis (Figure 1A) and correlated with Child-Pugh score (r_s = 0.484, P = 0.001) and its components bilirubin (r_s = 0.381, P = 0.009), albumin (r_s = -0.339, P = 0.021) and the international normalized ratio (INR, r_s = 0.445, P = 0.002) (Table 2). Because of lack of association of the IP index with creatinine, a correlation of the IP index with the MELD score failed the level of statistical significance (Figure 1B). There was no significant association the IP index with diabetes, serum creatinine or co-medication with proton pump inhibitors (PPI) or non-selective beta blockers (NSBB).

Open in New Tab   Full Size Figure   Download Figure

Figure 1 Intestinal permeability is impaired in cirrhosis and correlates with severity. A: Boxplots (according to Tukey’s definition with outliers) of the intestinal permeability (IP) index as determined by the lactulose/mannitol ratio in healthy controls and in patients with cirrhosis stratified for Child-Pugh stage; B: Scatter dot and non-parametric correlation (Spearman’s rho and P value) of IP index with MELD score.

Although there was no overall correlation of the IP index with LBP serum concentrations (Figure 2A), median LBP levels were significantly higher in patients with pathological IP than in patients without (median 25 μg/mL vs 13 μg/mL, P = 0.01). In addition, higher serum concentrations of IL-6 correlated with higher IP index (Figure 2B). When serum was assessed for concentrations of I-FABP, an enterocyte-specific marker of mucosal injury, I-FABP was higher than the expected normal range (210-870 pg/mL) in 72% (26/36) of patients. There was no evidence for a correlation of I-FABP with the IP index (Figure 2C) or with pathological IP (P = 0.61).

Open in New Tab   Full Size Figure   Download Figure

Figure 2 Correlation of intestinal permeability index with surrogates of inflammation and enterocyte mass. Scatter dot plots of the intestinal permeability (IP) index in cirrhotic patients with serum concentrations of (A) lipopolysaccharide-binding protein (LBP), (B) Interleukin-6 (IL-6) and (C) intestinal fatty acid binding protein (I-FABP). Spearman’s rho and P value for the non-parametric correlation are indicated.

Within 365 d follow-up, 78% (36/46) of the patients were hospitalized after a median of 42 d (range, 11-216 d) resulting in 111 hospitalizations with a total duration of 1464 d. On average 2 hospitalizations per patient with a median duration of 20 d were observed. Nineteen (41%) patients had a total of 33 episodes of hospitalization with admission for bacterial or fungal infection or development of infection during hospital stay. The most frequent infections were urinary tract infections in 9 (20%) patients, SBP in 7 patients (15%), spontaneous or secondary bacteremia/fungemia in 6 (13%) patients (4 × Gram-positive cocci, 1 ×E. coli, 1 ×Candida spp.), Clostridium difficile-associated enterocolitis in 5 (11%) patients. Less frequently observed infections were purulent skin and soft tissue infections including wound infections (n = 2), pneumonia (n = 1), campylobacter enteritis (n = 1), spontaneous bacterial empyema (n = 1), monomicrobial bacterascites (n = 1) and severe sepsis with unknown focus (n = 1). The median time to hospitalization for infection was 71 d (range, 17-302). Patients who developed bacterial infection had higher Child-Pugh score and higher MELD scores at baseline, lower platelet count and higher I-FABP serum concentration than patients who did not. The IP index at baseline was not associated with the occurrence of bacterial or fungal infection in general (Table 3). The IP index was not higher in patients who urinary tract infection (median 0.22 vs 0.15, P = 0.41), bacteremia/fungemia (0.16 vs 0.19, P = 0.71) or Clostridium difficile-associated enterocolitis (0.19 vs 0.18, P = 0.49) compared to patients who did not.

However, patients who developed SBP had a higher IP index (median 0.27) than patients who did not (median 0.14, P = 0.018). The accuracy of the IP index to identify patients with subsequent SBP [area under the ROC curve (AUROC) 0.780, P = 0.02] were comparable to the MELD score (AUROC 0.744, P = 0.04), Child-Pugh score (AUROC 0.736, P = 0.05), serum I-FABP (AUROC 0.879, P = 0.03) or the platelet count (AUROC 0.700, P = 0.10). When IP index was stratified according to tertiles, the cumulative 1-year incidences of SBP were 0%, 20% ± 13%, and 34% ± 12% (estimates ± standard error), in patients in the lowest, median and highest tertile according to Kaplan-Meier analysis (P for linear trend = 0.016) (Figure 3A). The cumulative incidences of hospitalization for infection did not significantly differ between patients in the lowest (28% ± 12%), median (43% ± 13%), and highest (63% ± 13%) tertile (P for linear trend = 0.096) (Figure 3B).

Open in New Tab   Full Size Figure   Download Figure

Figure 3 Cumulative incidence of bacterial infections. Kaplan-Meier curves displaying (A) the cumulative incidence of spontaneous bacterial peritonitis (SBP) (censored at transplant or death; event at date of SBP) and (B) the cumulative incidence of bacterial or fungal infections (censored at transplant or death; event at hospitalization with infection) according to the intestinal permeability (IP) index at baseline stratified by tertiles [lowest tertile: T1 (< 0.111), median tertile: T2 (0.111-0.226), highest tertile: T3 (> 0.226)]. Log-rank test for linear trends (T1 to T3) over strata is indicated.

Twenty-eight percent (13/46) of the patients died within 365 d of follow-up after a median of 116 d. Nine percent (4/46) underwent liver transplantation after a median of 180 d. One patient was lost to follow-up after 58 d, and three patients after 267-292 d. Among the 19 patients who developed bacterial infection, 9 (47%) died within the observation period, in contrast to 4 out of 27 (15%) patients who did not develop infection (P = 0.02).

We observed 2 (29%) deaths in patients with normal IP compared to 11 (28%) deaths in patients with pathologically increased IP (P = 1.000). There was no significant correlation between pathological IP and mortality (P = 0.82), transplant-free survival (P = 0.90) or infection-free survival (P = 0.90) according to Kaplan-Meier analysis. Overall survival and transplantation-free survival were lowest in patients in the median tertile group compared to patients in the highest and lowest tertile, and the transplantation-free survival did not differ between the groups (Figure 4).

Open in New Tab   Full Size Figure   Download Figure

Figure 4 Survival and event-free survival. Kaplan-Meier curves of (A) cumulative survival (censored at transplant; event at death), (B) transplant-free survival (events at death or transplants) and (C) infection-free survival (censored at transplant; events at hospitalization with infection or death) according to intestinal permeability (IP) index at baseline stratified by tertiles [lowest tertile: T1 (< 0.111), median tertile: T2 (0.111-0.226), highest tertile: T3 (> 0.226)]. Log-rank test for linear trends (T1 to T3) over strata is indicated.

In contrast to the IP (AUROC 0.550, P = 0.56), higher MELD score (AUROC 0.736, P = 0.006), higher Child-Pugh score (AUROC 0.736, P = 0.006), and higher I-FABP concentration (AUROC 0.774, P = 0.005) were significantly associated with 1-year infection-free survival when linearly assessed in ROC analysis and also in Cox regression after dichotomization. Higher serum IL-6 (AUROC 0.660, P = 0.06) showed a trend for predicting infection-free survival when assessed as linear continuous variable and was significant in a Cox regression model after dichotomization. In multivariate analysis only MELD score > 13 and the presence of ascites at baseline predicted infection-free survival (Table 4).

Increased intestinal permeability (IP) is a frequent phenomenon in patients with cirrhosis and has been linked to pathological bacterial translocation, bacterial infections and higher mortality mainly in retrospective studies. However, prospective studies confirming this association are scarce.

To employ pathological IP as an indicator and predictor of complications in cirrhosis in clinical practice, it is important to prospectively assess the association of IP with the development of bacterial infections and mortality in cirrhosis and to compare its predictive accuracy with other biomarkers of complications in cirrhosis.

In this prospective study, increased IP was associated with more advanced liver disease and increased systemic inflammation at baseline but failed to predict adverse clinical outcome in a 1-year follow-up period when focusing on the endpoints mortality, liver transplantation and infections.

Patients with ascites are at the highest risk of developing bacterial infections. For risk stratification, clinicians should apply well-established composite scores, such as Child-Pugh or the model for end-stage liver disease, and pro-inflammatory serum markers but not dual sugar tests of intestinal permeability.

Intestinal permeability refers to the barrier function of the gut, which - if compromised - allows potentially harmful substances to penetrate across this barrier.

This is well performed and interesting prospective clinical study for analyzing the impact of increased intestinal permeability on mortality and the occurrence of infection in patients with liver cirrhosis. The results are described and discussed in a clear way.