Intestinal permeability in human nonalcoholic fatty liver disease: A systematic review and meta‐analysis

Abstract Background The gut‐liver axis is considered to play a critical role in the development and progression of nonalcoholic fatty liver disease (NAFLD). The integrity of the epithelial barrier is crucial to protect the liver against the invasion of microbial products from the gut, although its exact role in NAFLD onset and progression is not clear. Methods We performed a systematic review and meta‐analysis of studies that addressed the intestinal permeability (IP) in association with NAFLD presence or severity as defined by the presence of nonalcoholic steatohepatitis (NASH) and the degree of steatosis, hepatic inflammation or fibrosis. A total of 14 studies were eligible for inclusion. Results Studies investigating IP in adult (n = 6) and paediatric (n = 8) NAFLD showed similar results. Thirteen of the included studies focussed on small IP, two studies on whole gut permeability and none on colonic permeability. In the pooled analysis, NAFLD patients showed an increased small intestinal permeability compared to healthy controls based on dual sugar tests (standardized mean difference 0.79, 95% CI 0.49‐1.08) and serum zonulin levels (standardized mean difference 1.04 ng/mL, 95% CI 0.40‐1.68). No clear difference in IP was observed between simple steatosis and NASH patients. Furthermore, whole gut and small intestinal permeability increased with the degree of hepatic steatosis in 4/4 studies, while no association with hepatic inflammation or fibrosis was observed. Conclusion Based on the limited number of studies available, IP appears to be increased in NAFLD patients compared to healthy controls and is associated with the degree of hepatic steatosis.


| INTRODUC TI ON
Nonalcoholic fatty liver disease (NAFLD) is the most common liver disease in the Western world in both adults and children. NAFLD prevalence is approximately 25% in the adult population and 8% in the paediatric population. 1,2 The spectrum of NAFLD ranges from nonalcoholic fatty liver (NAFL), nonalcoholic steatohepatitis (NASH) to liver fibrosis, cirrhosis and hepatocellular carcinoma. 1 To date, the exact pathophysiology of NAFLD has not completely been elucidated and it is not clear to what extent paediatric NAFLD differs from adult NAFLD. 3 The interaction between the gut and the liver, the so-called 'gutliver axis', is considered to play a critical role in development and progression of NAFLD in both children and adults. 4 Crosstalk between gut and liver is facilitated through the intestinal barrier. This intestinal barrier consists of structural elements (mucus and closely lined epithelial cells sealed by tight junctions), immune cells and soluble mediators (eg IgA, antimicrobial peptides). 4

An intact intestinal barrier is
able to restrict translocation of bacterial products, while allowing active transport from nutrients across the tight junctions. 4 The epithelial integrity of the intestinal barrier can be assessed in vivo by measuring the intestinal permeability (IP). Increased IP can lead to translocation of microbial products from the gut to the liver through the portal system.
Known factors that contribute to an increased IP include consumption of a Western diet (ie high fat intake), gut microbiome perturbations, pro-inflammatory cytokines, alcohol and use of antibiotics. 5,6 Currently, a number of non-invasive tests to measure IP in humans are being used. Urinary recovery of orally administered sugars (ie sucrose, lactulose to mannitol ratio (L/M), lactulose to rhamnose ratio (L/R), sucralose to erythritol ratio (S/E) and sucralose) are widely accepted as markers for IP. Five hour (h) urinary sucrose levels are used as indicator for gastroduodenal permeability, 5-6 h L/M and L/R as indicators for small intestinal permeability, and 5-24 h or 0-24 h S/E as indicators for colon and whole gut permeability respectively. By using the ratio of two sugars with different size and therefore different transport mechanism (paracellularly versus transcellularly), correction for differences in renal function, intestinal transit time and gastric emptying is possible. 7 Urinary recovery of a single substance cannot correct for these factors, which can differ between patients and thereby affect the outcome. Other substances used to measure IP in vivo are various polymers of polyethylene glycol (PEG) and 51 Cr-labelled ethylenediaminetetraacetic acid ( 51 Cr-EDTA). 7 More recently, zonulin, a 47-kDa protein, has been introduced as a potentially useful systemic marker for small intestinal and gastroduodenal permeability, but not for colon permeability. 8 Serum zonulin has emerged as a relevant biomarker because it is an important factor to regulating IP by modulating intercellular tight junctions. 9,10 However, the specificity of serum zonulin as biomarker for small intestinal permeability remains uncertain. 11 In both adults and paediatric NAFLD patients, several studies investigated IP and its role in the pathogenesis and progression from NAFL to NASH. [12][13][14][15][16][17] However, the exact association between IP and NAFLD severity (degree of steatosis, hepatic inflammation, fibrosis or presence of NASH) is not clear. The aim of this systematic review and meta-analysis was to summarize studies in humans on the association between in vivo IP alterations and NAFLD presence and/or severity. We hypothesize that IP is increased in NAFLD, being most pronounced in progressive disease as characterized by the presence of NASH, advanced steatosis, hepatic inflammation or hepatic fibrosis. Furthermore, in the included studies, we will summarize the clinical parameters (eg anthropometric data and blood biochemical variables), which have been observed to correlate with the degree of IP in NAFLD patients.

| ME THODS
Reporting of this systematic review and meta-analysis was performed according to the PRISMA guidelines (preferred reporting items for systematic reviews and meta-analyses). 18

| Search strategy
A systematic literature search was conducted in 2020 (week 38) in both PubMed and Embase. The following keywords, synonyms and MeSH terms were used: non*alcoholic fatty liver disease, Nonalcoholic Fatty Liver Disease, fatty liver disease, NAFL, Fatty Liver, Non-alcoholic Fatty Liver Disease, non*alcoholic steatohepatitis, NASH, nonalcoholic steatohepatitis, liver steatosis, hepatic steatosis, liver steatosis, intestinal barrier, gut barrier, gut permeability, permeability, zonulin. This resulted in 1070 hits and after exclusion of duplicates, 783 were included for screening of abstracts and full text. In addition, references of selected articles were assessed and included if suitable.

| Eligibility criteria
Studies on in vivo IP measurements in human NAFLD patients were included in this systematic review. Studies were eligible for inclusion when the following inclusion criteria were met: (i) original peer reviewed research paper in English, (ii) the study population or a subgroup of the population consist of NAFLD patients (diagnosed with liver biopsy or imaging) without cirrhosis (because cirrhosis Highlights • To date, the role of intestinal permeability (IP) in human NAFLD is not clear.
• Studies investigating IP in NAFLD mostly focus on small intestinal permeability.
• IP appears to be increased in NAFLD patients and appears to be positively associated with the degree of hepatic steatosis.
• IP is not associated with the degree of hepatic inflammation or fibrosis.
itself can lead to an increased IP), (iii) in vivo IP measurements (ie by urinary excretion of orally administered sugars, 51 Cr-EDTA or polyethylene glycol or by serum zonulin levels) and (iv) comparison of IP between groups (healthy controls (HC) vs NAFLD or NAFL vs NASH). Exclusion criteria included: (i) review articles, letter to the editor, commentaries, (ii) animal studies, (iii) studies investigating ex vivo permeability (ussing chambers) and solely microbial translocation via endotoxin/ lipopolysaccharide (LPS) levels.

| Selection process and data extraction
To reduce selection bias, all titles and abstracts were screened for eligibility (based on in-and exclusion criteria) independently by two authors (HV, TDM). After consensus full text of selected articles were again independently checked for eligibility by the same authors (HV, TDM). Furthermore, both authors independently extracted all data using standardized data extraction forms. Data on patient characteristics (HC and NAFLD), method of NAFLD diagnosis (imaging or biopsy), IP test, main outcome (IP comparison between groups and relationship to liver histology) and observed correlations between the degree of IP and clinical factors was extracted. In case of disagreement on eligibility, the two reviewers came to consensus after discussing the article with a third reviewer.

| Quality assessment
The methodological quality of the selected studies was assessed by two independent researchers (HV, TDM) using the Newcastle-Ottawa Quality Assessment Scale (NOS) for case-control studies. 19 The NOS-score was converted to (Agency for Healthcare Research and Quality) AHRQ standards using the following thresholds: good

| Statistical analysis
Meta-analyses were performed using a random effect model with If both a BMI matched (or obese) control group and a normal weight control group were available in one study, data of the BMI matched control group was used in the analysis. All data were entered as mean ± SD.
When the original results were only reported as median and (IQR) we estimated mean and SD using the formula proposed by Wan et al. 20 In the studies where the data were included in figures and not provided numerically, we used software program Plot Digitizer to extract data. The pooled standardized median difference with 95% CIs were presented in forest plots. Heterogeneity of study results was tested with χ 2 and I 2 calculations. We intended to assess publication bias by visual examination of the funnel plot and the Egger test for funnel plot asymmetry.

| Study selection
Twenty-eight studies were eligible for full text screening. Thirteen of 28 studies matched the criteria and were included in this review.
One additional study was identified through reference checking.
Excluded studies did not specify alcohol consumption (n = 3), did not use an in vivo IP test (n = 2), did not investigate the association between IP and NAFLD presence or severity (n = 5) or did not include a control group (HC or NAFL) (n = 5). Further details on the selection process can be found in the flowchart ( Figure 1).

| Study characteristics
Nine studies investigated IP with urinary recovery of orally administered molecules (ie sugars or 51 Cr-EDTA) ( Table 1) and five studies investigated IP with serum zonulin levels ( Table 2). Only two studies investigated whole gut permeability by means of 24 h urinary recovery of 51 Cr-EDTA or sucralose while all other studies focused on small intestinal permeability. 12,21 In five of fourteen studies BMI was not significantly different between the control group and NAFLD group.
However, in only two studies BMI matching of the control group with the NAFLD group (implemented in the study design) was performed. 14,22 Nine (5 adult and 4 pediatric) of fourteen studies used the golden standard, liver biopsy, to diagnose NAFLD, while the five other studies (one adult and four paediatric) used ultrasound. Study characteristics of all included studies are summarized in Table 1 (urinary recovery of orally administered molecules) and Table 2 (serum zonulin).   Overall NASH patients had no significantly different serum zonulin levels compared to NAFL patients (standardized mean difference 1.44 ng/ mL, 95% CI −0.13-3.00, I 2 = 95%) ( Figure 3B). When pooled separately ( Figure 3B), in both adult and paediatric patients, no difference in serum zonulin levels between NASH and NAFLD patients was observed. c Some degree of hepatocellular steatosis, and characteristic lobular mixed inflammation is sufficient to diagnose NASH.

| Whole gut permeability in NAFLD
d Steatosis and any inflammation is sufficient for the diagnosis of NASH.
or sucralose. 12,21 Data were not pooled because different markers were used. In the study of Farhadi et al, 24 h sucralose excretion was not significantly different between HC (n = 12), NAFL (n = 6) and NASH (n = 10) patients (Table 1). 21 In the study of

| Association between small intestinal permeability and NAFLD severity
Five of the included studies investigated the association between small intestinal permeability and one or more parameters of NAFLD severity (degree of steatosis, fibrosis, ballooning or inflammation) ( Table 3). The association between small intestinal permeability and the degree of hepatic steatosis was investigated in three studies (two paediatric and one adult) ( Table 3). [14][15][16] In all studies more advanced hepatic steatosis was associated with an increased small intestinal permeability. To quantify hepatic steatosis all studies used the histological NAFLD activity score (NAS) (Tables 1 and 2). The association between small intestinal permeability and hepatic fibrosis was investigated in four studies (three pediatric and one adult), [13][14][15][16] while three paediatric studies investigated the association with hepatic inflammation and hepatic ballooning (Table 3). 14

| Association between whole gut permeability and NAFLD severity
The association between whole gut permeability (24 h 51 Cr-EDTA) and NAFLD severity was investigated in one adult study. Miele et al observed that 24 h 51 Cr-EDTA excretion was significantly increased in NAFLD patients with moderate to severe steatosis (S2-3) compared to NAFLD patients with minimal or mild steatosis (S1). 12 Furthermore, no difference in degree of hepatic fibrosis, hepatic inflammation or ballooning was observed between patient with normal and increased 24 h 51 Cr-EDTA excretion. 12

| Factors that significantly correlated with IP in NAFLD patients
Four of the included studies (one adult and three paediatric) reported significant correlations between small intestinal permeability and clinical factors including anthropometric data and blood biochemical variables ( correlation between small intestinal permeability and the degree of insulin resistance. 14,22 In addition two other studies observed a positive correlation between small intestinal permeability and systemic LPS levels. 16,24 Other correlations ie with BMI, systolic blood pressure and blood ALT, IL-6, triglycerides, γ-GT and HDL-C levels are only observed in single studies (Table 4).  propionate were unaffected in faecal samples from paediatric NAFLD patients. 28,29 As IP was often investigated in a singular paediatric or adult study, subgroup analysis by age is not desirable. Therefore, future studies are needed to investigate differences in IP between adult and paediactric NAFLD patients.

| D ISCUSS I ON
Evidence is less convincing when comparing NAFL and NASH patients as investigated in six studies. In the pooled analysis, an increased small intestinal permeability was found by L/M (three studies) but not zonulin (three studies). It should be noted that results need to be interpreted with care because of substantial heterogeneity between studies for both parameters. Furthermore the number of study subjects in both adults or paediatric studies is very low. In previous studies, IP has also been associated with several metabolic abnormalities including obesity, dyslipidaemia and hyperglycaemia. 5 In NAFLD patients, increased IP is believed to induce hepatic steato-  Mechanistically, translocation of bacterial products (eg LPS), leads to activation of toll-like receptor 4 in the liver and results in hepatic inflammation and fibrogenesis. 4,6,28 The amount of NAFLD subjects with significant hepatic fibrosis or inflammation in the included studies is relatively low what may explain why no association between IP and hepatic fibrosis or inflammation was found. Furthermore, most studies included in this review focus on small intestinal permeability, while colon permeability was not investigated. Microbiome perturbation in the colon have been associated with NAFLD presence and severity and are believed to harm the integrity of the gut barrier. 28,30 In mice, high fat diet feeding has been observed to induce metabolic abnormalities, systemic and liver inflammation which was accompanied by an increased colon permeability. 31  Ohlsson et al suggest that serum zonulin might rather be a biomarker for low-grade inflammation than for IP, because zonulin is identical to prehaptoglobin-2, not enterocyte specific and associated with overweight, obesity and hyperlipidemia. 10,11,33,34 Furthermore, in the study of Linsalata et al serum zonulin did not correlate with the L/M but did correlate with serum IL-6 and serum IL-8 concentrations in 91 subjects (39 irritable bowel syndrome, 32 coeliac disease and 20 HC).

Association between IP and clinical variable
Finally, to date zonulin is the only known regulator of intestinal tight junction but it is likely that other zonulin unrelated pathways are also important in this process. Caution must be taken when using serum zonulin as a biomarker for small IP. Therefore, studies using zonulin as marker for small intestinal IP were analysed separately. 11 This systematic review has some limitations. Firstly, because of the observational nature of all included studies in this systematic review only associations and not causalities were investigated.
Secondly, substantial inter-study heterogeneity was noted in most analyses. In this review, only studies investigating in vivo IP by means of urinary excretion of orally administered substances or serum zonulin levels were included. Studies using circulating LPS levels, the major component of the outer membrane of Gramnegative bacteria, as marker for IP were not included as circulating LPS measurements are not site-specific and have a high false-positive rate. 35 Thirdly, because of the small number of studies included in the meta-analysis the presence of publication bias cannot be ruled out and subgroup analysis is not desirable. Finally, only 14 studies were included, which were small in terms of sample size, focused on both paediatric and adult NAFLD and most of them had poor quality.
In conclusion, small intestinal permeability appears to be increased in NAFLD patients compared to healthy controls and appears to be positively associated with the degree of hepatic steatosis. However, included studies where small in sample size, had poor quality and showed high heterogeneity. To date, no clear evidence is available that small intestinal or whole gut permeability increases with NAFLD severity (presence of NASH, hepatic inflammation or fibrosis). Future studies should also focus on colonic permeability in NAFLD patients.

CO N FLI C T S O F I NTE R E S T
All authors report no conflicts of interest relevant to this article.