High prevalence of celiac disease in autoimmune hepatitis: Systematic review and meta‐analysis

Previous studies investigating the prevalence of celiac disease (CD) in individuals with autoimmune hepatitis (AIH) have shown highly variable results. We therefore aimed to examine the prevalence of CD in individuals with AIH.


| INTRODUC TI ON
Autoimmune hepatitis (AIH) is a severe chronic and progressive autoimmune liver disorder characterized by immune-mediated inflammation of the liver that may proceed to cirrhosis and liver failure. [1][2][3] The incidence of AIH ranges between 0.85 and 1.68 per 100 000 person-years, predominantly afflicting women. [4][5][6] AIH may have a multifactorial pathogenesis, where both genetics, immune response regulation and environmental factors play a crucial role. 2,3,7 AIH have been linked to several extrahepatic autoimmune disorders, including autoimmune thyroiditis, diabetes, rheumatoid arthritis, inflammatory bowel disease and potentially also to celiac disease (CD). 8 CD is an immune-mediated enteropathy triggered by gluten in genetically susceptible individuals and is characterized by circulating celiac-specific autoantibodies and gastrointestinal symptoms including malabsorption caused by small intestinal villous atrophy but also extraintestinal symptoms. 9 In prior studies, the prevalence of CD has ranged from 2% to 20% in children and adults with AIH. [10][11][12] While the prevalence of CD in AIH individuals seems to be higher than in the general population (~1%), 9 prior studies have been small, mostly single-centre studies resulting in large variability in results and lack of precise estimates. [10][11][12] It is known that CD is associated with various types of liver diseases 13 and that the use of a gluten-free diet has been shown to improve liver injury in CD-associated hepatopathy, 14 autoimmune cholangitis 15 and non-alcoholic fatty liver disease (NAFLD). 16 Although the data is more sparse in those who have AIH, there is one report that indicated that, in individuals with AIH, the use of a gluten-free diet for CD improved liver injury, whereas the AIH treatment alone had not been satisfying in liver recovery. 17 However, it remains unknown whether CD worsens AIH activity or causes separate liver injury in addition to AIH.
Because of the highly varying CD prevalence data in AIH in earlier studies, we performed a meta-analysis and investigated the prevalence of CD in AIH.

| Eligibility criteria
After full-text review, publications with non-English language, poster presentations, conference abstracts, letters to editors, review articles, meta-analyses and publications without sufficient data were excluded ( Figure 1). For our main analysis, CD had to be biopsy-proven with villous atrophy (Marsh stage III; 8 studies). In a secondary analysis, a broader definition of CD was used, also including studies where CD was confirmed with at least one positive serological marker (tissue transglutaminase antibody (TTG), endomysial antibody (EMA) or antigliadin antibody (AGA)). The secondary analysis included a total of 15 articles.

| Data collection process and data items
The PRISMA guidelines 18 were followed, and relevant articles were reviewed in detail by LH and IG. From each article, publication date, country of origin, age of the individuals, CD definition and AIH definition (for further details, see Section 3), number of individuals with AIH, number of AIH individuals with CD, number of controls, number of controls with CD, number with a positive serology, number with a biopsy-proven CD, Marsh grade (III) for the definition of CD and study design, were retrieved.

| Statistics
In order to reduce the influence of smaller studies on the summary estimate, a fixed effect model was used for the calculation of the weighted prevalence of CD in individuals with AIH. 19 The prevalence was stated with 95% confidence intervals. The heterogeneity was reported with I 2 in percentage, and a P-value < 0.05 was considered statistically significant. For detection of potential publication bias, a meta-funnel analysis was carried out (Appendix, Figure S1). The quality of each study was evaluated and commented in Section 3. STATA 13 was used for all statistical analyses.

Study highlights What is known
• Autoimmune hepatitis is a chronic and progressive autoimmune liver disorder.
• Earlier studies have investigated the prevalence of celiac disease in individuals with autoimmune hepatitis, but with highly varying results.

What is new here
• Individuals with autoimmune hepatitis demonstrate a higher prevalence of celiac disease compared to the general population.
• Individuals with autoimmune hepatitis constitute a risk group for concurrent celiac disease.
• Testing for celiac disease could be of great value in patients with autoimmune hepatitis.

| Patient and public involvement
Patients or the public were not involved in the design, or conduct, or reporting, or dissemination plans of our research.
To be eligible for our main analysis (n = 8), we required that the authors stated that Marsh III (villous atrophy) was needed for the CD diagnosis. Diagnosis criteria for AIH in each paper are listed in Table 1. AIH was diagnosed in the majority of papers using either the original 1993 score, 51

| Prevalence of biopsy-verified CD in AIH
The 8 identified studies for our main analysis 10

| Prevalence of CD, defined through biopsy or positive serology, in AIH
When allowing cases of CD defined either through biopsy (Marsh III) 10

| Risk of bias across studies
A meta-funnel plot (Appendix, Figure S1) detected signs of publication bias, in that smaller studies were more likely to be published if they showed higher prevalence of CD in AIH than if they showed a lower prevalence.

| D ISCUSS I ON
Our meta-analysis revealed that the pooled prevalence of biopsyverified CD in individuals with AIH was 3.5% (95% CI = 1.6%-5.3%) and 2.9% (95% CI = 2.1%-3.8%) if combining both biopsy-or serologyproven CD (95% CIs for the two definitions overlapped). Our main biopsy-and serology-proven CD. 29 In addition, it was based on only 3 publications, 2 of which we excluded as these did not meet our inclusion criteria (Caprai 30 : insufficient data; Diamanti 20 : letter to editor, Figure 1). The shared publication included in both meta-analyses was the paper by El-Shabrawi et al. 38 The 3.5% pooled prevalence of biopsy-proven CD in individuals with AIH was significantly higher than the overall prevalence of around 1% of CD seen in most general populations. 9 AIH and CD may share some gene coding for class II human leukocyte antigens, which may explain their concurrent existence. 27 Volta et al 11  Another study has also suggested that gluten removal may significantly improve liver injury, where AIH treatment alone had not given satisfying results. 17 This effect could be due to an increased intestinal permeability with both circulating and residual tissue transglutaminase antibodies in the liver, modifying self-antigens causing liver injury. 27 Current guidelines on AIH from the American Association for the Study of Liver Diseases recommend screening AIH individuals for CD at diagnosis. 2 Ours is the largest meta-analysis that includes both children and adults, and the results support the statement of celiac screening in the AIH guidelines and additionally provide a more accurate estimate of the prevalence of CD in AIH. However, the guidelines do not state specifically how the screening for CD should be done. 2 An initial serology-based CD screening could possibly be used as opposed to esophagogastroduodenoscopy with biopsy given that the prevalence of CD using both definitions is very similar with overlapping CIs (biopsy: 3.5% (95% CI = 1.6%-5.3%) vs combining biopsy and serology: 2.9% (95% CI = 2.1%-3.8%)). This would most likely be a more cost-effective and less invasive method for CD screening. CD screening may also possibly be done more than once, not only at AIH diagnosis, as concurrent autoimmune disorders, including CD, could develop over time. 8

F I G U R E 2
Prevalence of biopsy-verified celiac disease (Marsh III) in autoimmune hepatitis. The confidence interval was calculated for each study using ±2 SD for the purpose of the inverse weighting of each study. Thus, individual 95% lower confidence interval may have a negative value. Confidence intervals for individual studies therefor have no true meaning. Legend: The 95% CIs for the individual studies in the figure were calculated by us and only used to assign weight for the meta-analysis. For that reason, the lower 95% CI may be negative below 0 This high rate of CD diagnosis seen in our results could be an underestimate of the prevalence of CD in AIH. This is because the treatment of AIH (steroids, prednisone, budesonide and immuno-  Interestingly, the pooled prevalence of biopsy-verified CD was higher in our study than that of serology-or biopsy-verified CD diagnosis, 3.5% (95% CI = 1.6%-5.3%) vs 2.9% (95% CI = 2.1%-3.8%), although the latter is a more liberal set of criteria. Since serologybased CD tends to be more prevalent than biopsy-verified CD (however often preceded by a positive serology), our results do not indicate that serology-positive CD is less common, only that studies based on serological markers found on average lower CD prevalence. Of note, the 95% CI for the two outcomes were overlapping.

| Strengths and limitations
We included AGA in our search terms for serology-based CD since that antibody was used in older studies of CD; however, AGA may overestimate the prevalence of CD due to false-positive cases. 55 Additionally, we cannot rule out that publication bias has impacted on our results since smaller studies may have been published more often if they showed a higher prevalence of CD (Appendix, Figure S1). Finally, our study had limited power in sub-analyses, and we were unable to calculate, for instance, area-specific prevalence estimates of CD. In a meta-analysis by Singh et al, 56 the prevalence of CD varied between 0.4% in South America and 0.8% in Europe and Oceania (general population), and we cannot rule out that CD is both more and less prevalent in AIH populations in different parts of the world.

| Clinical implications
The meta-analysis found a high prevalence of CD in patients with AIH. Patients with AIH may benefit from CD screening.

| Conclusion
This meta-analysis of 567 individuals with AIH revealed a CD diagnosis in every 30 AIH individuals.

ACK N OWLED G EM ENTS
The authors thank Karolinska Institutet University Library for conducting the search for publications and JFL -the Swedish Research Council and Karolinska Institutet -for grants and financial support.

CO N FLI C T O F I NTE R E S T
Dr Ludvigsson coordinates a study on behalf of the Swedish IBD quality register (SWIBREG). This study has received funding from Janssen Corporation. All other authors declare that they have no conflict of interest and nothing to declare.

D I SCL A I M ER
This manuscript represents the views of the authors.

FU N D I N G A N D RO LE O F TH E FU N D I N G S O U RCE S
Karolinska Institutet. The funders did not influence the study or the decision to submit the study; all authors have completed the ICMJE uniform disclosure form at www.icmje.org/coi_discl osure.pdf.

D E TA I L S O F E TH I C S A PPROVA L
Not relevant.