Patients With Inflammatory Bowel Disease on Treatment Have Lower Rates of Celiac Disease

Abstract

Background

The association between celiac disease and inflammatory bowel disease (IBD) has been studied; however, the impact of IBD therapy on celiac disease is not known. Using a large database, we sought to describe the association of celiac disease and IBD and the impact of IBD treatment.

Methods

We queried a large multicenter database (Explorys Inc.), an electronic health record data aggregate from 26 American health care systems. We identified a cohort of patients with celiac disease and IBD between 1999 and 2020 and conducted a statistical analysis using a multivariate model.

Results

Of the 72,965,940 individuals in the database, 133,400 had celiac disease (0.18%), 191,570 (0.26%) had ulcerative colitis (UC), and 230,670 (0.32%) had Crohn disease (CD). Patients with IBD were more likely to have a diagnosis of celiac disease (odds ratio [OR], 13.680), with a greater association with CD. Treated patients with UC and with CD, respectively, had a lower risk association with celiac disease compared to those not undergoing IBD treatment, specifically corticosteroids (OR, 0.407 and 0.585), 5-aminosalicylates (OR, 0.124 and 0.127), immunomodulators (OR, 0.385 and 0.425), and anti-tumor necrosis factor drugs (OR, 0.215 and 0.242). There was no lower risk association in the vedolizumab group, but there was a higher risk association among the ustekinumab group.

Conclusions

In this large dataset, we showed a bidirectional association between celiac disease and IBD that was stronger with CD. Patients with IBD treated using corticosteroids, 5-aminosalicylates, immunomodulators, or anti-tumor necrosis factor drugs had a lower association with celiac disease. Additional studies are required to determine the underlying mechanisms for IBD therapy–related modification of celiac disease incidence.

Introduction

Celiac disease and inflammatory bowel disease (IBD) are characterized by chronic immune-mediated inflammation of the gastrointestinal tract.^{1, 2} Both are complex diseases with genetic and environmental factors contributing to dysregulation of the immune response, leading to chronic inflammation. Although it is well-known that certain human leukocyte antigens (HLA-DQ2 and HLA-DQ8) are associated with celiac disease,³ studies have not shown a higher prevalence of an HLA-celiac disease genotype in patients with IBD compared to the general population.⁴ However, genome studies have observed that celiac disease and IBD share non-HLA genes.^5-7 Furthermore, environmental factors are better understood in celiac disease, where gluten ingestion is the main putative factor.⁸

Celiac disease is considered one of the most frequent immune-mediated diseases, with an overall prevalence rate in the United States ranging between 0.7% and 1%.^9-11 Diagnosed IBD has a lower prevalence, with rates ranging from 25 to 200 per 100,000 in Crohn disease (CD) to 35 to 250 per 100,000 in ulcerative colitis (UC).⁸ The clinical presentations of celiac disease and IBD can be similar in that patients may present with abdominal pain, alteration in bowel function (commonly diarrhea), and weight loss. Frequently, patients with celiac disease or IBD may present with extraintestinal manifestation, systemic involvement, and other autoimmune diseases.

There are several studies that have reported the association between celiac disease and IBD; a few have suggested a bidirectional relationship, including a recent meta-analysis.¹² Using a large database, we sought to evaluate the association between celiac disease and IBD and describe the impact of different IBD medications on celiac disease.

METHODS

Database

A retrospective analysis was conducted using a multicenter research platform engine developed by IBM Watson Health (Explorys Inc., Cleveland, OH)¹³: a large, inclusive, American-based population database that provides data from the electronic medical records of >70 million patients from 26 major health care centers in the United States. Participating health care institutions are granted access to the deidentified patient data. The Systematized Nomenclature of Medicine—Clinical Terms (SNOMED CT) hierarchy is used to arrange diagnoses, findings, and procedures,¹⁴ and RxNorm is used to map prescription drug orders.¹⁵ The Explorys research platform provides an interactive engine to browse clinical data, generate multiple cohorts based on diagnoses, and investigate temporal relationships. It is Health Insurance Portability and Accountability Act-compliant because Explorys utilizes deidentified data. The database research engine rounds population counts to the nearest 10 and treats all counts between zero and 10 as equivalent to protect patient confidentiality. The database has been used previously in multiple medical fields including gastroenterology, psychiatry, pediatrics, and surgery.^16-21

Patient Selection

The study cohort (IBD cohort) was identified by searching the database for a SNOMED CT diagnosis of “Crohn’s disease” or “ulcerative colitis” between 1999 and 2020. Subsequently, patients with SNOMED CT diagnoses of “celiac disease” after at least 365 days of “Crohn’s disease” and “ulcerative colitis” diagnosis were identified by selecting the option “after at least” in Explorys. Likewise, another study cohort was identified for “celiac disease” patients and individuals with SNOMED CT diagnoses of “Crohn’s disease” and “ulcerative colitis” after at least 365 days of “celiac disease” by selecting the option “after at least” in Explorys. By using the option of “first ever” occurrence, only sentential episodes were included in the study cohort. The control group was then identified as patients without SNOMED CT diagnoses of “Crohn’s disease” or “ulcerative colitis” for the first cohort and “celiac disease” for the second cohort. We contrasted our selected SNOMED CT terms for UC, CD, and comorbidities to equivalent ICD-9-CM codes (Supplementary Table 1), which have been utilized in prior studies using population-based data to identify their respective conditions.^{22, 23}

Covariates

Age, sex, and race-based data were collected. Potential confounding factors were also identified including age, sex, race, tobacco smoking, and medications including corticosteroids (prednisone, prednisolone, methylprednisolone, budesonide), 5-aminosalicylates (5-ASAs), immunomodulators (azathioprine, 6-mercaptopurine, methotrexate), and biological agents (infliximab, adalimumab, certolizumab, vedolizumab, ustekinumab). Nonsteroidal anti-inflammatory drug (NSAID) prescriptions were also captured (ibuprofen, diclofenac, ketorolac, COX-2 inhibitors).

Statistical Analysis

There were 4 major cohorts: patients with IBD, patients without IBD, patients with celiac disease, and patients without celiac disease. Patients with IBD were compared to patients without IBD, and patients with celiac disease were compared to patients without celiac disease. The overall period prevalence rates were calculated by dividing the total number of individuals with CD, UC, and celiac disease separately by the total number of individuals in Explorys, respectively. In the first 2 cohorts (IBD vs non-IBD), the prevalence rates of celiac disease in the study (IBD) and control groups (non-IBD) were calculated by dividing the total number of individuals with celiac disease over the total number of individuals with IBD and without IBD, respectively. Likewise, the same was replicated for the other cohorts (celiac disease vs non-celiac disease): The prevalence rates of CD and UC among the study (celiac disease) and control group (non-celiac disease) were calculated by dividing the total number of patients with CD and UC over the total number of patients with celiac disease and non-celiac disease, respectively.

To adjust for potential confounding from the aforementioned covariates, 8192 searches were performed to account for every probability. All statistical analyses were performed using IBM SPSS Statistics version 25, with celiac disease diagnosis as an outcome between the IBD group and the propensity score-matched control patients without IBD and correction for multiple comparisons.

RESULTS

Descriptive Epidemiology

Baseline characteristics of patients with IBD cohort and control and celiac disease vs control group are displayed in Table 1 and Table 2. Of the 72,965,940 individuals in the database, 133,400 had celiac disease (0.18%), 191,570 (0.26%) had UC, and 230,670 (0.32%) had CD. A new diagnosis of celiac disease after 1 year of IBD diagnosis between 1999 and 2020 was 1.59% and 0.90% in patients with CD and with UC, respectively, compared with 0.16% in patients without IBD (P < 0.0001). A new diagnosis of CD and UC in patients with celiac disease between 1999 and 2020 was 2.75% and 1.11%, respectively, compared with 0.29% and 0.25%, respectively, in the non–celiac disease population (P < 0.0001). A new diagnosis of IBD and celiac disease among patients with microscopic colitis was 10.5% and 2.6%, respectively, and a new diagnosis of microscopic colitis among patients with celiac disease was 0.01%.

Multivariate Regression Analysis of Study Cohorts

In the multivariate model, after adjusting for age, sex, race, smoking, and aforementioned medications, there was a significant risk association between IBD and celiac disease (odds ratio [OR], 13.680; 95% confidence interval [CI], 13.454-13.909; P < 0.0001). The risk association of celiac disease was higher with CD (OR, 24.473; 95% CI, 23.981-24.974; P < 0.0001) compared with UC (OR, 5.686; 95% CI, 5.686-5.979; P < 0.0001).

Effect of Medications on Risk of Developing Celiac Disease Among Patients With IBD

Using multivariate analysis to adjust for confounding factors, we evaluated the effect of medications on the risk of developing celiac disease among patients with IBD. Medications included corticosteroids (prednisone, prednisolone, methylprednisolone, budesonide), 5-ASAs, immunomodulators (azathioprine, 6-mercaptopurine, methotrexate), and biologic agents including anti-tumor necrosis factor (TNF) drugs (infliximab, adalimumab, certolizumab), vedolizumab, and ustekinumab. Fig. 1 outlines the prevalence of immunomodulators and biologic agents among patients with concomitant celiac disease and IBD based on category. Patients treated using 5-ASAs had the lowest risk association between celiac disease and IBD compared with other biologics; the risk association was 0.124 (95% CI, 0.117-0.132) for UC and 0.127 (95% CI, 0.120-0.134) for CD. Compared with patients who did not receive anti-TNFs, patients who received infliximab, adalimumab, or certolizumab had a lower risk association with celiac disease: 0.215 (95% CI, 0.198-0.234) for UC and 0.242 (95% CI, 0.227-0.258) for CD. Similar results were found for patients treated using azathioprine, 6-mercaptopurine, or methotrexate compared with those who did not receive an immunomodulator: 0.385 (95% CI, 0.350-0.423) for UC and 0.425 (95% CI, 0.396-0.456) for CD. Whereas there were no statistically significant associations with vedolizumab, patients with IBD who were treated using ustekinumab had a higher risk association with celiac disease. Patients who received corticosteroids had a lower risk association compared with those who were not treated using corticosteroids (Figs. 2, 3).

Discussion

In this large study, we showed a strong bidirectional association between celiac disease and IBD. In fact, after adjusting for confounding factors, a significant risk association was found between celiac disease and IBD (OR, 13.680; 95% CI, 13.454-13.909, P < 0.0001). The risk was higher with CD (OR, 24.473; 95% CI, 23.981-24.974; P < 0.0001) than with UC (OR, 5.686; 95% CI, 5.686-5.979; P < 0.0001), consistent with prior reports from outside the United States.^{12, 24} Interestingly, patients with IBD were less likely to be associated with celiac disease when they were treated using corticosteroids, 5-ASAs, immunomodulators, or anti-TNFs. Among those who received vedolizumab, there were no statistically significant associations, whereas there were increased risk associations among those who received ustekinumab.

Previous reports have shown a possible association between IBD and celiac disease. Two recently published large meta-analyses indicated a link between celiac disease and IBD in a bidirectional fashion, with one increasing the risk of the other. Pinto-Sanchez et al¹² analyzed 65 studies and showed a 4-fold increased risk of celiac disease in patients with IBD and a 10-fold increased risk of IBD in patients with celiac disease. Similarly, among 27 studies with a total of 41,482 patients, Shah et al²⁴ showed that celiac disease was associated with a more than 11-fold increased risk of developing IBD when compared to the risk in the non–celiac disease population. The risk was higher in CD (OR, 10.38; 95% CI, 4.53-23.80) compared with UC (OR, 6.98; 95% CI, 4.75-10.24). In contrast, patients with IBD were at a 2-fold increased risk of developing celiac disease compared with the general population. This finding is consistent with our findings, where we found a higher risk association with CD. Overall, this result suggests a strong link between celiac disease and IBD in a bidirectional fashion; however, the risk seems to be higher among patients with celiac disease to develop IBD rather than vice versa. Whether celiac disease is pathogenically linked to IBD or the association is linked to shared immunogenetics is not known.^{7, 25, 26}

There are several factors believed to contribute to this association. Both diseases are multifactorial, with genetic, environmental, and immunologic changes contributing to inflammation and disease. Even though studies have not shown a higher prevalence of HLA-DQ2 and HLA-DQ8 in patients with IBD compared with the general population,^{4, 27} genome studies have observed that celiac disease and IBD share non-HLA genes and an overexpression of major histocompatibility complex class I chain-related gene A.⁷ Furthermore, this trend has been reflected in studies showing a familial association between IBD and celiac disease.^{28, 29} In terms of environmental contributions, common factors like infections and changes in the microbiome are thought to play a role in celiac disease and IBD immunopathogenesis.^{1, 8} The hygiene hypothesis has been proposed mainly for IBD and to a lesser extent for celiac disease to explain the rising prevalence of these diseases in industrialized countries.⁸

The interactions between environmental factors and genetic alterations are reflected in common cellular and humoral immune reactions in both diseases. Intraepithelial T lymphocytes have a pivotal role in the pathogenesis of both diseases. Moreover, some autoantibodies are shared between both diseases, including anti-Saccharomyces cerevisiae antibodies^30-32 and antinuclear antibodies.^{33, 34} The shared genetics and potential common environmental factors may explain the higher risk association between celiac disease and IBD.^{7, 25, 26} However, based on the meta-analyses conclusions by Pinto-Sanchez et al and Shah et al,12, 24 the risk of developing IBD in patients with celiac disease is much higher than the risk of developing celiac disease in patients with IBD. Mucosal inflammation as a result of gluten exposure in patients with celiac disease may initiate a sequence of events that ultimately increases the risk of IBD. By increasing intestinal permeability and allowing unprocessed antigens to interact with the mucosal immune system or altering the microbiome, celiac disease may increase the risk of IBD, although this possibility is speculative. Furthermore, NSAIDs have been linked to several toxicities, including increasing the risk of intestinal mucosal and vascular permeability; in our cohort, the prevalence of NSAIDs prescriptions for patients with IBD were comparable with those for the control group.³⁵ Other factors that have been linked to increased intestinal permeability include stress, which is increased among patients with celiac disease.^{35, 36}

It is important to highlight that microscopic colitis has been associated with both IBD and celiac disease. In a population-based cohort study from Sweden, microscopic colitis was associated with a 17-fold increased risk of IBD.³⁷ Furthermore, in another prospective cohort of 1000 patients with celiac disease, Green et al³⁸ showed a 70-fold increased risk of developing microscopic colitis among patients with celiac disease. Similarly, we found an increased risk association between microscopic colitis and celiac disease (OR, 17.78; 95% CI, 11.34-27.89; P < 0.0001).

Despite the well-known association between IBD and celiac disease, there are limited data regarding the outcomes of patients and the natural course of having coexisting diseases. In a case-control study on 51 patients with coexisting IBD and celiac disease, Oxford et al³⁹ showed that patients with UC who had celiac disease were more likely to have pancolitis, whereas the coexistence of celiac disease among patients with CD did not influence its natural course. This finding was also observed in a case series in which 4 of 5 studied patients with UC and celiac disease developed pancolitis.⁴⁰

There are limited data in the literature regarding the effects of immunomodulators and biologic agents on the risk of developing celiac disease among patients with IBD. Suppression of gluten-specific T-cell mediated immune responses through antibody-based therapies using anti-TNFs has shown histologic improvements in 2 case studies with refractory celiac disease.^{41, 42} In a retrospective study by Bengi et al,⁴³ none of the patients with IBD who developed celiac disease were receiving biologic agents. In our study, patients with IBD undergoing treatment were less likely to have celiac disease compared to those who were not undergoing treatment. Patients with IBD who received an anti-TNF agent (infliximab, adalimumab, or certolizumab) had a lower risk association with celiac disease compared with those who did not receive anti-TNFs. Furthermore, patients who received azathioprine, 6-mercaptopurine, or methotrexate had lower rates of having celiac disease than those who did not receive any immunomodulator. Thiopurines have been used in refractory celiac disease; however, there are no studies examining the effect of thiopurines on the risk of developing celiac disease in IBD. The lower risk association between IBD and celiac disease among those receiving an immunomodulator could represent a protective effect of the medication. In fact, tissue transglutaminase and endomysial antibodies have decreased posttransplantation without gluten elimination in patients with end-stage liver disease, likely as a result of the use of immunosuppression.⁴⁴

Interestingly, patients with IBD who received 5-ASAs also had a lower risk association with celiac disease. Although our study cannot confer a direct treatment effect of 5-ASAs on celiac disease, 5-ASAs do have anti-inflammatory and antioxidant effects that may restore the oxidative balance in patients with celiac disease.⁴⁵ Studies have shown that mesalamine decreases a gluten-induced cytokine response in vitro, and it has been used in some refractory condition in vivo.⁴⁶ In a small study including 4 patients treated using small-intestinal-release mesalamine and 6 patients who received small-intestinal-release mesalamine and oral budesonide for refractory celiac disease, 50% had complete response and an additional 10% had partial response.⁴⁷ In our study, patients who were treated using 5-ASAs for their IBD had significantly lower rates of celiac disease compared to patients with IBD who did not receive treatment.

Corticosteroids are also associated with lower rates of celiac disease. The use of corticosteroids in celiac disease was first described in 1978 and has been a common treatment for patients with refractory disease.⁴⁸ Small studies have shown symptomatic and histologic improvement in patients taking systemic corticosteroids and locally active corticosteroids.^{49, 50} More recent studies have reported the clinical usefulness of oral budesonide, including open-capsule protocol, to minimize systemic adverse effects.^{51, 52} In a small randomized clinical trial, treatment using daily budesonide (6 mg) for a duration of 4 weeks resulted in increased body weight and a decreased number of bowel movements among patients with refractory celiac disease compared with those who followed a gluten-free diet alone. Well-being scores were higher in patients treated using both a gluten-free diet and budesonide compared with those following a gluten-free diet alone.⁵³ Although we did not study the rates of new IBD diagnoses in nonresponsive celiac disease, this frequency should be a consideration in patients lacking improvement despite a gluten-free diet.

There are limitations to this study that are inherent to claims database analyses. A potential bias is the classification of celiac disease. For example, serologic assays for celiac disease can be falsely positive in patients with IBD, and villous atrophy can be seen in patients with IBD, both sources of potential misclassification. Another limitation is in the diagnoses of celiac disease and IBD that are SNOMED CT–coded and lack objective validation, eg, endoscopic, radiographic, and pathological confirmation; disease severity; and duration of therapy. Both IBD and celiac disease have a genetic basis, and family history and extent would enhance our understanding of this interesting association; however, the Explorys database does not include discrete information on family history. Another limitation is the possibility of overdiagnosing celiac disease in IBD and vice versa because of the higher rate of endoscopic evaluation and biopsies in patients with IBD and celiac disease compared to the general population. We also noted that medical comorbidities (such as hypertension, diabetes mellitus, and obesity) were more common among patients in our cohort. Patients with IBD are known to have several comorbidities related to their disease process along with medication adverse effects.^54-56 Furthermore, patients with IBD receive more clinical attention, and hence they may undergo more testing that can potentially lead to detection bias.^{57, 58} Multiple steps were taken to avoid confounding bias, which is usually inherent to large database studies. Despite these limitations, this is the largest database study to evaluate the association of celiac disease and IBD and the potential impact of IBD therapy on celiac disease incidence. The large size effect of the association between celiac disease and IBD is unlikely to be affected by the limitations of a claims database.

Conclusions

There is an association between celiac disease and IBD, with CD having a higher association compared with UC. Patients with IBD treated using 5-ASAs, immunomodulators, or anti-TNFs have less association with celiac disease compared with those not on treatment. Celiac disease should be considered in patient with IBD with unexplained anemia or refractory IBD. Conversely, patients with nonresponsive celiac disease should also be evaluated for IBD in the appropriate clinical context.

Abbreviations

Abbreviations
 
• 5-ASA

  5-aminosalicylates

 
• CD

  Crohn disease

 
• CI

  confidence interval

 
• HLA

  human leukocyte antigen

 
• IBD

  inflammatory bowel disease

 
• NSAIDs

  nonsteroidal anti-inflammatory drugs

 
• OR

  odds ratio

 
• SNOMED CT

  Systematized Nomenclature of Medicine—Clinical Terms;

 
• TNF

  tumor necrosis factor;

 
• UC

  ulcerative colitis

Author contributions: study conception and design: Alkhayyat, Abou Saleh, Rubio-Tapia, Regueiro; acquisition of data: Alkhayyat, Almomani, Abou Saleh, Abureesh, Mansoor; analysis and interpretation of data: Alkhayyat, Abou Saleh, Abureesh, Zmaili, Regueiro; drafting of manuscript: Alkhayyat, Almomani, Abureesh, Abou Saleh, Zmaili; critical revision: El Ouali, Rubio-Tapia, Regueiro; statistical analysis: Alkhayyat, Abou Saleh, Abureesh, Mansoor; study supervision: Rubio-Tapia, Regueiro; guarantors of the article: Rubio-Tapia, Regueiro; approval of final draft: all authors.

References