Risk of incident pancreatitis in patients with celiac disease: A population-based matched retrospective cohort study

Abstract

BACKGROUND

Celiac disease (CD) is an autoimmune disorder associated with an increased risk of pancreatitis, yet large-scale studies examining long-term risk and specific etiologies in CD patients are scarce.

AIM

To assess the long-term risk of pancreatitis in CD patients.

METHODS

We conducted a population-based cohort study with consecutive patients diagnosed with CD using the TriNeTx research network. Each patient was matched to a patient in the control group using a 1:1 propensity score matching to minimize confounding effects. The primary outcomes were the incidence of acute pancreatitis and chronic pancreatitis, and the secondary outcome was to assess the etiologies of pancreatitis. The incidence was estimated using a Cox proportional hazards model with a hazard ratio (HR) and 95% confidence interval (CI).

RESULTS

A total of 160228 patients were identified to have CD, and the remaining 250725 individuals without CD were considered as controls. At 7-year follow-up, CD patients exhibited a significantly higher risk of acute pancreatitis (HR = 2.05; 95%CI: 1.93-2.17) and chronic pancreatitis (HR = 1.42; 95%CI: 1.31-1.54) compared to controls. Elevated risks for alcohol-induced (HR = 1.35), biliary (HR = 1.37), and idiopathic pancreatitis (HR = 1.49) were also observed. Findings remained robust across all follow-up intervals and sensitivity analyses.

CONCLUSION

Patients with CD have a substantially increased long-term risk of acute and chronic pancreatitis, including alcohol-related, biliary, and idiopathic subtypes. These findings support the routine surveillance of pancreatitis in CD management and highlight the need for further research into disease-specific risk factors and mitigation approaches.

Key Words: Celiac disease; Pancreas; Pancreatitis; Acute pancreatitis; Chronic pancreatitis; Pancreatitis etiology; Population-based study

Core Tip: Celiac disease (CD) is an immune-mediated disorder with well-recognized gastrointestinal and extraintestinal manifestations. However, limited large-scale studies have evaluated its association with pancreatic disease. We investigated the long-term risk of acute and chronic pancreatitis in a large, propensity score-matched cohort of patients with CD. CD was associated with a significantly higher risk of pancreatitis, including alcohol-related, biliary, and idiopathic subtypes, compared with matched non-CD controls. This risk persisted for at least seven years. Our findings highlight the importance of pancreatitis surveillance and preventive strategies as an integral part of CD management.

INTRODUCTION

Celiac disease (CD) is a chronic, immune-mediated disorder precipitated by the ingestion of gluten-derived peptides, which trigger an aberrant immune response in genetically susceptible individuals[1]. Current estimates indicate that CD affects approximately 0.5% to 1% of the global population, with an increasing prevalence noted in recent years, particularly in the United States[2-5]. This rise has been attributed to heightened vigilance in screening and the advent of more sensitive and specific diagnostic modalities, such as serologic assays and duodenal biopsies[6,7]. Extraintestinal complications of CD, especially pancreatic complications, are a major area of concern because of their severity and impact on quality of life[8-10]. The growing incidence of CD brings new challenges, notably an expanded spectrum of extraintestinal complications. Among these, pancreatic manifestations - including acute pancreatitis (AP) and chronic pancreatitis (CP) - pose significant clinical concern due to their impact on patient morbidity and quality of life[8-10]. Global epidemiological data reveal a steady increase in the prevalence of AP and CP, with CP rising by 52.1 cases per 100000 individuals between 1996 and 2016[11]. Although gallstones and alcohol misuse remain primary etiological factors for pancreatitis, emerging evidence implicates CD as an independent risk factor for both AP and CP[9,12-15]. Previous studies linking CD and pancreatitis have generally been limited by small sample sizes, single-center designs, and short observation periods, reducing their generalizability[14,15].

A recently published large-scale cohort study demonstrated significantly higher rates of pancreatitis among patients with CD compared to controls, supporting the hypothesis of an association between CD and pancreatic disease[15]. However, robust, population-based investigations evaluating long-term outcomes remain sparse. Therefore, the present study aims to address this gap by systematically examining the long-term risk of AP and CP, as well as their specific etiologies, among individuals with CD in a large, multicenter cohort. By leveraging comprehensive, real-world data, the present study aimed to investigate the long-term risk of AP and CP among CD patients, as well as to identify the frequency of various etiologies of pancreatitis in this population and to clarify disease associations and inform clinical decision-making for patient management.

MATERIALS AND METHODS

Methods

A population-based, multicenter retrospective cohort study was performed using data from the TriNetX research network (Cambridge, MA, United States), a federated platform that aggregates anonymized electronic health records from diverse healthcare organizations nationwide. Data integrity and quality were ensured prior to dataset inclusion via systematic extraction protocols and validation checks. Patients were included using standard International Classification of Diseases [classification, Ninth and Tenth Revision, Clinical Modification (ICD-9 and ICD-10-CM)] and Current Procedural Terminology classification terminologies. As a federated network, TriNetX obtained a waiver of review and consent from the Western Institutional Review Board by utilizing solely aggregated counts and statistical summaries of deidentified information.

Study participants

The cohort comprised all patients with a diagnosis of CD between March 2005 and April 2022, identified using validated ICD codes. Details of the codes, selection algorithm, and quality controls are provided in the Supplementary material. For the control group, individuals without CD from the same time frame were selected from the general population. Individuals with prior diagnoses of pancreatic or gallbladder disorders, those who had undergone procedures such as endoscopic retrograde cholangiopancreatography (ERCP) or cholecystectomy, and patients with any malignancy, including pancreatic cancer, before the index date (defined as the CD diagnosis date or equivalent reference date for controls) were excluded (Figure 1).

Figure 1 Flow chart of patient selection for study cohorts. ERCP: Endoscopic retrograde cholangiopancreatography; CD: Celiac disease.

Matching process

Each participant with CD was paired with a control from the non-celiac cohort through 1:1 propensity score matching (PSM) to minimize potential confounding. The propensity score model incorporated predefined covariates recognized as possible confounders, including age, sex, race or ethnicity (Hispanic, non-Hispanic White, non-Hispanic Black, or non-Hispanic other), nicotine dependence, alcohol-related conditions, body mass index, type 2 diabetes mellitus, hypertension, hyperlipidemia, chronic lower respiratory diseases (chronic obstructive pulmonary disease or asthma), ischemic heart disease, heart failure, pulmonary heart disease, cerebrovascular disease, chronic kidney disease, fatty liver, hepatic cirrhosis, malignant neoplasms, additional autoimmune disorders (type 1 diabetes, autoimmune thyroiditis, rheumatoid arthritis, or inflammatory bowel disease), gallbladder disease, pancreatic disorders, familial hypercholesterolemia, hypercalcemia, and procedures known to elevate pancreatitis risk (ERCP and cholecystectomy). Laboratory variables, including serum cholesterol, low-density lipoprotein, triglycerides, and hemoglobin A1c, were also included in the model (Table 1).

Table 1 Baseline characteristics of patients with celiac disease and without celiac disease before and after propensity score matching, n (%).

Variables	Before propensity matching			After propensity matching
	CD (n = 160228)	Non-CD (n = 250725)	SMD	CD (n = 137685)	Non-CD (n = 137685)	SMD
Age, mean ± SD	39.1 ± 22	47.2 ± 18.9	0.3959	42.9 ± 21.1	43.8 ± 20.2	0.0462
Age by categories
< 18 years	33862 (21.1)	18691 (7.5)	0.3985	18650 (13.5)	15394 (11.2)	0.0300
18 years to < 40 years	49079 (30.6)	18691 (7.5)	0.0825	42014 (30.5)	43884 (31.9)	< 0.0001
40 years to < 60 years	43112 (26.9)	67446 (26.9)	0.2202	42978 (31.2)	43884 (31.9)	0.0142
≥ 60 years	34175 (21.3)	93062 (37.1)	0.1670	34043 (24.7)	34523 (25.1)	0.0081
Gender
Female	113373 (70.8)	182852 (72.9)	0.0483	99033 (71.9)	100461 (72.9)	0.0223
Ethnicity
Hispanic or Latino	7885 (4.9)	22051 (8.8)	0.0602	7463 (5.4)	6616 (4.8)	0.0279
Race
White	131627 (82.2)	169727 (67.7)	0.3382	110190 (80.0)	110461 (80.2)	0.0049
Black or African American	6385 (4.0)	41009 (16.4)	0.4181	6385 (4.6)	6306 (4.6)	0.0027
Asian	1795 (1.1)	5213 (2.1)	0.0765	1743 (1.3)	1633 (1.2)	0.0073
Others	19865 (12.4)	33341 (13.3)	0.0765	18860 (13.7)	18805 (13.7)	0.0012
BMI (kg/m2), mean ± SD	25.8 ± 7.4	28.8 ± 7.3	0.4132	26.6 ± 7.2	28 ± 7.4	0.1974
Nicotine dependence	9288 (5.8)	27951 (11.1)	0.1931	9118 (6.6)	8873 (6.4)	0.0072
Alcohol related disorders	2801 (1.7)	8911 (3.6)	0.1126	2777 (2.0)	2672 (1.9)	0.0055
Comorbidities
Hyperlipidemia	16210 (10.1)	50025 (20.0)	0.2778	16152 (11.7)	16568 (12.0)	0.0093
Type 2 diabetes mellitus	13489 (8.4)	32412 (12.9)	0.1464	11912 (8.7)	12108 (8.8)	0.0015
Hypercholesterolemia	7367 (4.6)	18577 (7.4)	0.1186	7139 (5.2)	7394 (5.4)	0.0083
Hypertriglyceridemia	1397 (0.9)	3712 (1.5)	0.0565	1348 (1.0)	1488 (1.1)	0.0101
Hypercalcemia	1072 (0.7)	3313 (1.3)	0.0658	1046 (0.8)	1087 (0.8)	0.0034
Autoimmune diseases
Type 1 diabetes mellitus	6694 (4.2)	6943 (2.8)	0.0770	4590 (3.3)	2862 (2.1)	0.0774
Rheumatoid arthritis	2569 (1.6)	17670 (7.0)	0.2701	2565 (1.9)	2530 (1.8)	0.0019
Gallbladder diseases
Cholecystitis	1312 (0.8)	2525 (1.0)	0.0198	1181 (0.9)	1150 (0.8)	0.0025
Cholelithiasis	3241 (2.0)	8548 (3.4)	0.0854	3138 (2.3)	3048 (2.2)	0.0044
Other diseases of the gallbladder	1598 (1.0)	3271 (1.3)	0.0284	1483 (1.1)	1423 (1.0)	0.0043
Other diseases of the biliary tract	1520 (0.9)	3121 (1.2)	0.0284	1426 (1.0)	1374 (1.0)	0.0038
Family history
Familial hypercholesterolemia	100 (0.1)	295 (0.12)	0.0044	100 (0.1)	68 (0.0)	0.0094
Procedure
ERCP	407 (0.3)	896 (0.4)	0.0256	393 (0.3)	344 (0.3)	0.0013

CD: Celiac disease; BMI: Body mass index; ERCP: Endoscopic retrograde cholangiopancreatography; SMD: Standardized mean difference.

Propensity scores for all participants in both cohorts were generated using logistic regression applied to the input matrices. The analysis was conducted in Python version 3.6.5 (Python Software Foundation), utilizing standard libraries such as NumPy and sklearn. To verify the consistency of results, parallel analyses were conducted in R version 3.4.4 (R Foundation for Statistical Computing, Vienna, Austria). A greedy nearest-neighbor algorithm with a caliper of 0.1 pooled standard deviations was used for matching. Balance was assessed using standardized mean differences, with an a priori threshold of 0.10 for acceptable balance. The order of the rows in the covariate matrix was randomized before matching to further mitigate bias.

Study outcomes

The primary outcomes were the incidence of AP and CP among CD patients. Secondary outcomes included the incidence of specific etiologic subtypes: Biliary, alcohol-induced, and idiopathic pancreatitis. Definitions for each outcome utilized validated ICD-9/10 codes, as detailed in the Supplementary material.

Statistical analysis

All statistical analyses were performed in real-time using the TriNetX platform. Continuous data were described as mean ± SD or as median and interquartile range, depending on the distribution. Categorical data were reported as counts and percentages. Cox proportional hazards regression was used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for each outcome. Proportionality was verified analytically, and these analyses were performed using the survival package in R version 3.2.3. Baseline-adjusted HRs were calculated for all comparisons. Results were validated by cross-checking outputs from SAS v9.4 statistical software (SAS Institute). Statistical significance was defined by a two-sided alpha ≤ 0.05.

Secondary and sensitivity analysis

A prespecified secondary analysis assessed time-dependent risk for pancreatitis, estimating HRs at distinct follow-up intervals, specifically, initiating follow-up at 1 and 3 years post-CD diagnosis. Sensitivity analyses further evaluated the robustness of results by excluding cases diagnosed within the first year after CD diagnosis, accounting for potential diagnostic misclassification and disease heterogeneity.

RESULTS

Baseline characteristics

This study analyzed data from a total of 410953 individuals. Of these, 160228 had a documented history of CD and were assigned to the primary cohort. The remaining 250725 participants, who had no prior diagnosis of CD, formed the comparison group. Detailed baseline characteristics for those in the CD cohort are presented in Table 1. The majority of CD patients were white (n = 131627, 82.2%) and female (n = 113373, 70.8%). 49079 (30.6%) patients were between the ages of 18 years and 40 years. Rates of common comorbidities were lower in the CD group compared to the non-CD patients, including hyperlipidemia (n = 16210, 10.1%), type 2 diabetes mellitus (n = 13489, 8.4%), and familial hypercholesterolemia (n = 100, 0.1%). Notably, type 1 diabetes prevalence was higher in the CD cohort (n = 6694, 4.2%). Substance use measures, such as nicotine dependence (5.8%) and alcohol-related disorders (1.7%), were less frequent among CD patients. Parallel demographic and comorbidity trends were observed in the control group, with a higher proportion of individuals ≥ 60 years (37.1%) and higher rates of metabolic and cardiovascular comorbidities. Furthermore, although slightly higher in the non-CD cohort, both cohorts had a similar percentage of cholecystitis, gallbladder disease, and biliary tract disease. There was also a similar percentage of patients with a history of ERCP procedures in the CD and non-CD cohorts.

Clinical characteristics

Results of laboratory findings for CD and non-CD patients are presented in Supplementary Table 1. Serologic and laboratory findings revealed CD-associated immunoreactivity: Serum gliadin peptide immunoglobulin A and immunoglobulin G were markedly elevated in CD patients compared to controls (Immunoglobulin A: 22 units/volume vs 7.75 units/volume; immunoglobulin G: 25.7 units/volume vs 7.6 units/volume). Several liver function tests for CD patients were lower compared to those for non-CD patients, including alanine aminotransferase (29.2 U/L vs 31.8 U/L), aspartate aminotransferase (28.8 U/L vs 30.7 U/L), total bilirubin (0.557 mg/dL vs 0.593 mg/dL), and gamma-glutamyl transferase (54.3 U/L vs 93.4 U/L). However, CD patients had higher levels of alkaline phosphatase (99.2 U/L vs 88.4 U/L) and albumin (4.14 g/dL vs 4.02 g/dL) compared to the non-CD cohort. Markers of inflammation, specifically sedimentation rate and C-reactive protein, were reduced in CD patients vs controls. No statistically significant differences were detected for lipid parameters.

Study outcomes

At the 1-year follow-up, 2483 CD patients and 848 non-CD patients developed AP, while 1288 CD and 861 non-CD patients developed CP Patients with CD had a significantly higher risk of AP (HR = 2.94; 95%CI: 2.72-3.18), CP (HR = 1.50; 95%CI: 1.37-1.63) (Figure 2), alcohol-induced pancreatitis (HR = 2.19; 95%CI: 1.48-3.25), biliary pancreatitis (HR = 2.10; 95%CI: 1.56-2.84), and idiopathic pancreatitis (HR = 2.62; 95%CI: 1.86-3.69) (Figure 2) compared to non-CD patients. At the 3-year follow-up, 2978 CD and 1273 non-CD patients developed AP, while 1303 CD and 878 non-CD patients developed CP In comparison to non-CD patients, CD patients continued to have a higher risk of AP (HR = 2.35; CI: 2.20-2.50), CP (HR = 1.48; CI: 1.36-1.62) (Figure 3), alcohol-induced pancreatitis (HR = 1.66; CI: 1.21-2.29), biliary pancreatitis (HR = 1.52; CI: 1.21-1.92), and idiopathic pancreatitis (HR = 1.93; CI: 1.48-2.52).

Figure 2 Forest plot of hazard ratios for alcoholic, biliary, and idiopathic pancreatitis in the celiac disease cohort vs the non-celiac cohort (years 1-7). CI: Confidence interval.

Figure 3 Incidence of pancreatitis in celiac disease vs non-celiac disease cohorts. A: Incidence of acute pancreatitis in celiac disease vs non-celiac disease cohorts; B: Incidence of chronic pancreatitis in celiac disease vs non-celiac disease cohorts. HR: Hazard ratio; CI: Confidence interval; CD: Celiac disease.

At the 5-year follow-up, 3219 CD patients and 1527 non-CD patients developed AP. 1430 CD and 1003 non-CD patients developed CP. The higher risk of AP (HR = 2.13; CI: 2.00-2.26), CP (HR = 1.43; CI: 1.32-1.55) (Figure 3), alcohol-induced pancreatitis (HR = 1.44; CI: 1.08-1.92), biliary pancreatitis (HR = 1.45; CI: 1.19-1.79), and idiopathic pancreatitis (HR = 1.67; CI: 1.33-2.10) continued in CD patients compared to non-CD patients. At the 7-year follow-up, 3345 CD and 1661 non-CD patients developed AP, while 1501 CD and 1069 non-CD patients developed CP. CD patients had a higher risk of AP (HR = 2.05; CI: 1.93-2.17), CP (HR = 1.42; CI: 1.31-1.54), alcohol-induced pancreatitis (HR = 1.35; CI: 1.03-1.77), biliary pancreatitis (HR = 1.37; CI: 1.12-1.66), and idiopathic pancreatitis (HR = 1.49; CI: 1.21-1.84) continued in CD patients compared to non-CD patients.

Sensitivity analysis

The findings remained robust in sensitivity analyses, which initiated follow-up one year post-CD diagnosis, thereby excluding events occurring within the first year after the index date. The HR associated with CD was comparable in magnitude to that observed in the primary analysis (Figure 3).

DISCUSSION

Our multi-institutional population-based study shows that patients with CD have a significantly higher risk of developing AP and CP as compared to non-CD patients. Furthermore, this increased risk persisted throughout the seven-year study period, remaining elevated even with long-term follow-up. Interestingly, our subgroup analyses demonstrate that CD patients are at a significantly higher risk of developing alcohol-, gallstone-, and idiopathic pancreatitis in comparison to those without CD. Several pathophysiological mechanisms may explain our findings of an increased risk of AP and CP in patients with CD. Dysregulation (including lower circulating levels) in the signaling, secretion, and function of enteric hormones, such as secretin and cholecystokinin, associated with CD may result in gallbladder motility and delayed release of pancreatic enzymes, causing a functional outflow obstruction promoting the development of AP[8,16,17]. Low post-prandial cholecystokinin levels in untreated CD patients also impair pancreatic secretions[18]. Papillary stenosis from duodenal inflammation in CD may further reduce ampullary outflow[19]. Data from transgenic mice expressing human leukocyte antigen-DQ8 (a major histocompatibility complex antigen associated with CD) also suggest that genetically susceptible individuals with CD may be more vulnerable to developing AP and CP through activation of Th-1-associated cytokines[20,21].

Our findings are consistent with several previous studies. Autopsy studies in patients with CD exhibit pancreatic fibrosis and acinar atrophy - pathological hallmarks of CP, pointing towards a history of recurrent AP episodes, ultimately resulting in CP[20]. A large United States-based cohort study found CD patients at a higher risk for AP [odds ratio (OR) = 2.66; 95%CI: 2.55-2.77] and CP (OR = 2.18; 95%CI: 2.04-2.34) within at least 30 days of CD diagnosis[15]. Data from Sweden also showed an association between CD and pancreatitis [any pancreatitis OR; HR = 2.85; 95%CI: 2.53-3.21; CP: OR = 2.76 (95%CI: 2.36-3.22)][13]. The higher hazard ratios, as compared to our findings, may be explained by differences in the population, as well as the detailed PSM in our analysis, minimizing confounding. In the inpatient setting, Osagiede et al[14] found a significantly higher prevalence of AP in CD patients as compared to the non-CD population. Furthermore, they reported significantly lower morbidity and mortality from AP in the CD cohort, possibly resulting from a milder form of AP. Kumar et al[22] found that 27% of CD patients undergoing endoscopic ultrasound had findings consistent with CP. It is noteworthy that the results of our secondary outcomes are also supported by previous evidence. In their Swedish cohort, Sadr-Azodi et al[13] concluded that relative to non-CD patients, patients with CD had a higher incidence of gallstone-related and non-gallstone-related pancreatitis. Risk of idiopathic pancreatitis, the most common etiology among CD patients, was significantly higher (OR = 1.54; 95%CI: 1.34-1.77) as compared to non-CD patients[15].

To the best of our knowledge, this is the first study reporting a higher risk of AP and CP in CD patients from a large population-based dataset with a long follow-up period. Furthermore, our finding of an increased risk of alcohol-related pancreatitis among CD patients (despite accounting for common alcohol-related comorbidities, e.g., cirrhosis) has never been reported previously. Our large sample size and detailed data in TriNetX allowed for a robust analysis to detect the association between CD and pancreatitis, including a subgroup analysis. The multicenter nature of the TriNetX database facilitated a representative study population from centers across the United States, increasing the generalizability of our findings. A PSM analysis helped ensure well-matched cohorts, strengthening the validity of our findings. From a clinical and policy standpoint, these findings underscore the importance of emphasizing AP and CP prevention measures, particularly in patients newly diagnosed with or actively managing CD.

Strengths and limitations

This study offers several notable strengths. It features a well-defined control group and a thorough organ-specific evaluation centered on the pancreas. Adjustments were made for a wide range of lifestyle-related risk factors, as well as baseline variables and potential confounders that could influence the observed relationship between CD and pancreatitis risk. The large sample size used in the propensity score-matched analysis contributed to narrow CIs and enabled the detection of a substantial number of outcome events, thereby reinforcing the validity of our conclusions. Additionally, the inclusion of sensitivity analyses further supports the robustness of our findings.

However, the study is not without limitations. Its retrospective nature and reliance on electronic medical record data introduce the possibility of selection bias due to documentation and coding inaccuracies. Moreover, residual confounding from factors such as alcohol consumption, smoking habits, medication use, and other lifestyle variables cannot be entirely excluded. Notably, patients included in the study, as part of their recruitment, have a new or current diagnosis of CD; therefore, our findings on the incidence and risk of pancreatic complications may not be generalizable to patients with treated CD. Given the retrospective design of this study, it was not possible to investigate the underlying pathophysiological mechanisms that may explain the observed link between CD and pancreatitis. Furthermore, data on villous atrophy and adherence to a gluten-free diet were not included in the analysis. Future research is needed to determine whether villous atrophy - and its persistence over time - contributes to an elevated risk of pancreatitis. Similarly, it remains to be seen whether normalization of villous architecture or strict compliance with a gluten-free diet can reduce this potential risk. Future studies should address these gaps by incorporating prospective designs, biomarker analysis, and detailed dietary data to clarify the pathways and develop preventive strategies.

CONCLUSION

In summary, our study showed that patients with CD have a significantly increased risk of developing acute and CP, including alcohol-related, biliary, and idiopathic subtypes, over at least seven years of follow-up. These findings underscore the importance of vigilant pancreatitis surveillance, targeted preventive interventions, and additional research into CD-specific risk factors and management strategies. However, further research into risk factors associated with the development of pancreatitis in CD patients is warranted.