Association of Bile Duct and Gallbladder Diseases With the Use of Incretin-Based Drugs in Patients With Type 2 Diabetes Mellitus

Importance  The use of dipeptidyl-peptidase–4 (DPP-4) inhibitors and glucagon-like peptide 1 (GLP-1) analogues—a group of drugs used in the management of type 2 diabetes mellitus—may be associated with an increased risk of bile duct and gallbladder disease. To date, no observational study has assessed this possible association.

Objective  To determine whether the use of DPP-4 inhibitors and GLP-1 analogues is associated with an increased risk of incident bile duct and gallbladder disease in patients with type 2 diabetes.

Design, Setting, and Participants  A population-based cohort study linked the United Kingdom Clinical Practice Research Datalink with the Hospital Episodes Statistics database, yielding a cohort of 71 369 patients, 18 years or older, initiating an antidiabetic drug (including oral and injectable agents) between January 1, 2007, and March 31, 2014.

Exposures  Current use of DPP-4 inhibitors and GLP-1 analogues (alone or in combination therapy) compared with current use of at least 2 oral antidiabetic drugs.

Main Outcomes and Measures  Time-dependent Cox proportional hazards models were used to estimate hazard ratios (HRs) with 95% CIs of incident bile duct or gallbladder events (cholelithiasis, cholecystitis, cholangitis) causing hospitalization, comparing current use of DPP-4 inhibitors and GLP-1 analogues with current use of at least 2 oral antidiabetic drugs.

Results  During 227 994 person-years of follow-up, 853 of the 71 369 patients were hospitalized for bile duct and gallbladder disease (incidence rate per 1000 person-years, 3.7; 95% CI, 3.5-4.0). Current use of DPP-4 inhibitors was not associated with an increased risk of bile duct and gallbladder disease compared with current use of at least 2 oral antidiabetic drugs (3.6 vs 3.3 per 1000 person-years; adjusted HR, 0.99; 95% CI, 0.75-1.32). In contrast, the use of GLP-1 analogues was associated with an increased risk of bile duct and gallbladder disease compared with current use of at least 2 oral antidiabetic drugs (6.1 vs 3.3 per 1000 person-years; adjusted HR, 1.79; 95% CI, 1.21-2.67). In a secondary analysis, GLP-1 analogues were also associated with an increased risk of cholecystectomy (adjusted HR, 2.08; 95% CI, 1.08-4.02).

Conclusions and Relevance  The use of GLP-1 analogues was associated with an increased risk of bile duct and gallbladder disease. Physicians should be aware of this potential adverse event when prescribing these drugs.

Incretin-based drugs, which include dipeptidyl peptidase–4 (DPP-4) inhibitors and glucagon-like peptide 1 (GLP-1) analogues, are relatively new antidiabetic therapies recommended as second- to third-line treatments in patients with type 2 diabetes mellitus.¹ The DPP-4 inhibitors (eg, linagliptin, sitagliptin phosphate monohydrate, vildagliptin, and saxagliptin hydrochloride) and GLP-1 analogues (eg, exenatide and liraglutide) both induce activation of the GLP-1 receptor, which stimulates insulin secretion in a glucose-dependent fashion while also inhibiting secretion of glucagon.²

In addition to its role in glucose regulation, GLP-1 also has been shown to enhance the proliferation and functional activity of cholangiocytes.^3,4 This finding raises concern that the use of incretin-based drugs may increase the risk of bile duct and gallbladder diseases, such as cholecystitis and cholangitis, and possibly bile duct cancer. According to the World Health Organization event database, the use of incretin-based drugs has been reported in 1069 cases of bile duct or gallbladder disease (eg, cholelithiasis, cholecystitis, and cholangitis) and 79 cases of bile duct or gallbladder malignant neoplasm worldwide since 2007.⁵ Furthermore, in a recent randomized clinical trial, high-dose liraglutide (a GLP-1 analogue) for weight loss was associated with an increase in gallbladder-related events (eg, cholelithiasis, cholecystitis, and cholecystectomies) vs placebo.⁶ Although these outcomes are rarely life-threatening or involve serious complications, they are painful and often necessitate hospital admissions, resulting in costly diagnostic and therapeutic procedures. To our knowledge, no observational study has been conducted to assess this possible association. Such observational studies could provide much-needed safety information in the natural setting of clinical practice.

To address this issue, we conducted a population-based cohort study to determine whether the use of DPP-4 inhibitors and GLP-1 analogues is associated with an increased risk of bile duct and gallbladder disease in patients with type 2 diabetes.

Box Section Ref ID

This study was conducted by linking the United Kingdom (UK) Clinical Practice Research Datalink (CPRD) with the Hospital Episodes Statistics (HES) database. The CPRD records patient-level information from more than 13 million patients, representing approximately 8% of the UK population. Patients included in the CPRD have been shown to be representative of the UK population in terms of age and sex.^7,8 The data collected in the CPRD have been shown to be of high quality⁹ and include anthropometric information (eg, body mass index [calculated as weight in kilograms divided by height in meters squared]), lifestyle variables (eg, smoking and alcohol use), medical information (eg, diagnoses and procedures [coded using the Read code classification²⁷]), and prescriptions written by general practitioners (coded according to the UK Prescription Pricing Authority Dictionary²⁸).

The HES database contains all inpatient and day case hospital admission information, including primary and secondary diagnoses (coded using the International Statistical Classification of Diseases and Health-Related Problems, Tenth Revision [ICD-10]) and hospital-related procedures. The linkage of the HES database to the CPRD is possible from April 1, 1997, onward, and is limited to English general practices that have consented to the linkage scheme (currently representing 75% of all English practices).⁸

The study protocol was approved by the Independent Scientific Advisory Committee of the CPRD (protocol number 15_235) and by the Research Ethics Board of the Jewish General Hospital, Montreal, Quebec, Canada. Patient informed consent was not necessary since the data were anonymized for research purposes.

We first assembled a base cohort of patients newly treated for type 2 diabetes consisting of all patients 18 years or older initiating a new noninsulin antidiabetic drug (metformin, sulfonylureas, prandial glucose regulators, thiazolidinediones, acarbose, DPP-4 inhibitors, and GLP-1 analogues) between April 1, 1998, and December 31, 2013. We excluded patients initially treated with insulin (as these patients are likely to have advanced type 2 diabetes) and women with a history of polycystic ovarian syndrome and gestational diabetes (other known indications for metformin). All patients were required to have at least 1 year of medical history recorded in the CPRD before their first noninsulin prescription.

Using the base cohort, we assembled a study cohort composed of all patients who initiated a new antidiabetic drug class on or after January 1, 2007 (the year the first incretin-based drug was licensed in the United Kingdom). These patients included those newly treated with a noninsulin antidiabetic drug, as well as those who switched to or added on an antidiabetic drug from a class not previously used in their treatment history. Cohort entry was defined by the date of this new prescription. We excluded patients previously diagnosed with bile duct and gallbladder diseases (ICD-10 codes listed in eTable 1 in the Supplement), primary biliary cirrhosis (ICD-10 code, K74.3), biliary cancer (ICD-10 codes, C22.1, C23, C24, D37.6, and D01.5), human immunodeficiency virus or AIDS or use of highly active antiretroviral therapy,¹⁰ and hemolytic disorders (eg, sickle cell disease or hereditary spherocytosis)^11,12 at any time before cohort entry. We also excluded patients who used ursodeoxycholic acid or ursodiol and women who were pregnant in the year before cohort entry.

All patients were followed up until an inpatient or day case hospital admission for a bile duct– or gallbladder-related condition (eg, cholelithiasis, cholecystitis, cholangitis, gallstone pancreatitis, and other bile duct and gallbladder disorders, as the primary diagnosis for admission; ICD-10 codes listed in eTable 1 in the Supplement), or censored after receipt of a pregnancy-related diagnosis, end of registration with the general practice, last date of HES data availability, death from any cause, or end of the study period (March 31, 2014), whichever occurred first.

A time-dependent exposure definition was used, in which each person-day of follow-up was classified in 1 of the following 4 mutually exclusive categories: current use of a DPP-4 inhibitor (linagliptin, saxagliptin, sitagliptin, or vildagliptin; alone or in combination with other antidiabetic drugs), current use of a GLP-1 analogue (exenatide or liraglutide; alone or in combination with other antidiabetic drugs), current use of 2 or more oral antidiabetic drugs, and other antidiabetic drugs and treatment combinations (which included patients concurrently exposed to DPP-4 inhibitors and GLP-1 analogues [n = 1102]). For all categories, exposed person-time was defined by duration of 1 prescription plus a 30-day grace period. Thus, patients were considered continuously exposed if the duration of 1 prescription overlapped with the date of the next prescription, using the grace period in the event of 2 nonoverlapping successive prescriptions. Since incretin-based drugs are recommended as second- to third-line treatments in the management of type 2 diabetes,¹³ the reference category for all analyses consisted of current use of 2 or more oral antidiabetic drugs.

All models were adjusted for the following potential confounders measured at cohort entry: age, sex, year of cohort entry, body mass index (<25, 25-29.9, ≥30), smoking status (current, past, or never), and hemoglobin A_1c (HbA_1c) level (last measurement before cohort entry). The models were also adjusted for any of the following that occurred at any time before entry into the cohort: alcohol-related disorders (based on diagnoses related to alcoholism, alcoholic cirrhosis of the liver, alcoholic hepatitis, and hepatic failure), diabetic arterial complications (retinopathy, neuropathy, nephropathy, peripheral arteriopathy, myocardial infarction, and ischemic stroke), and history of gastric or small-intestine bypass surgery. Finally, we adjusted for the number of physician visits and use of morphinomimetics (which increase pressure of the sphincter of Oddi) and fibrates (which inhibit acetyl-coenzyme A acetyltransferase activity) in the year before cohort entry.

Cox proportional hazards regression models using the time-varying exposure definition were used to estimate adjusted hazard ratios (HRs) with 95% CIs of hospitalization for bile duct or gallbladder disease associated with current use of GLP-1 analogues and DPP-4 inhibitors, compared with current use of 2 or more oral antidiabetic drugs. All models were adjusted for the potential confounders listed above.

We performed 3 secondary analyses. First, we assessed whether incretin-based drugs were associated with hospitalizations resulting in cholecystectomy. Second, we evaluated whether the risk varied according to duration of continuous use among current users of incretin-based drugs (≤180, 181-360, and >360 days). Linear trend was assessed by including these duration categories as a continuous variable in the model. Finally, we determined whether duration of treated type 2 diabetes modified the association by including an interaction term between duration of treated diabetes (<5 vs ≥5 years) and the relevant incretin-based drug exposure categories.

We conducted 6 sensitivity analyses to assess the robustness of our findings. First, we repeated the primary analysis using grace periods of 0 and 90 days between successive prescriptions. Second, we repeated the analyses after excluding patients who used insulin (a marker of advanced diabetes) before cohort entry and censoring after receipt of a new prescription during follow-up. Third, we additionally adjusted the analyses for the use of antidiabetic drugs (metformin, sulfonylureas, thiazolidinediones, insulins, and other agents) in the year before cohort entry. Fourth, we repeated the analyses using multiple imputation for variables with missing information, such as body mass index, HbA_1c level, and smoking status. Fifth, we used the disease risk score approach as alternate means to control for confounding¹⁴; this approach has been shown to have a comparable performance as the propensity score method (eMethods in the Supplement).¹⁵ Finally, to explore the possibility of ascertainment bias, we performed an analysis excluding and censoring on pancreatic-related events (acute and chronic pancreatitis, congenital defects of the pancreas, and pancreatectomy). All analyses were conducted with SAS, version 9.4 (SAS Institute).

The cohort included 71 369 new users of antidiabetic drugs (Figure 1). The mean (SD) follow-up was 3.2 (2.0) years, for a total of 227 994 person-years. Overall, 853 patients had an incident hospital admission for bile duct or gallbladder disease during follow-up, corresponding to an overall incidence rate per 1000 person-years of 3.7 (95% CI, 3.5-4.0). Diagnoses for these admissions included (non–mutually exclusive categories) cholelithiasis (n = 563), cholecystitis (n = 368), cholangitis (n = 5), and other bile duct and gallbladder diseases (n = 151). Cholecystectomy was performed in 75 patients (8.8%).

Table 1 presents the characteristics of patients who received DPP-4 inhibitors, GLP-1 analogues, and other antidiabetic drugs at cohort entry. Compared with users of other antidiabetic drugs, users of GLP-1 analogues were younger, more likely to be obese, and more likely to have previously used sulfonylureas or insulin. Overall, patients who used DPP-4 inhibitors and GLP-1 analogues had longer durations of treated diabetes, higher HbA_1c levels, and were more likely to have microangiopathic diabetic arterial complications compared with users of other antidiabetic drugs. The median (first and third quartiles) durations of use for DPP-4 inhibitors and GLP-1 analogues were 365 (156-692) and 289 (139-538) days, respectively.

Table 2 presents the results related to the use of DPP-4 inhibitors. Compared with current use of 2 or more oral antidiabetic drugs, current use of DPP-4 inhibitors was not associated with an increased risk of bile duct and gallbladder disease (3.6 vs 3.3 per 1000 person-years; adjusted HR, 0.99; 95% CI, 0.75-1.32). Likewise, the risk did not vary with the duration of use, with all HRs around the null value (Table 2) or when stratifying on duration of treated diabetes (eTable 2 in the Supplement). Finally, current use of DPP-4 inhibitors was not associated with hospitalization resulting in a cholecystectomy (adjusted HR, 1.13; 95% CI, 0.67-1.93).

Table 3 presents the results related to the use of GLP-1 analogues. Compared with current use of 2 or more oral antidiabetic drugs, current use of GLP-1 analogues was associated with a 79% increased risk of bile duct and gallbladder disease (6.1 vs 3.3 per 1000 person-years; adjusted HR, 1.79; 95% CI, 1.21-2.67). Duration of treated diabetes did not modify the association (eTable 3 in the Supplement). In secondary analyses, the use of GLP-1 analogues for less than 180 days was associated with an increased risk of bile duct and gallbladder disease (adjusted HR, 2.01; 95% CI, 1.23-3.29) while longer durations of use were not statistically associated with an increased risk, although these duration categories were based on fewer events (Table 3). Finally, current use of GLP-1 analogues was also associated with an increased risk of undergoing a cholecystectomy (adjusted HR, 2.08; 95% CI, 1.08-4.02).

The results of the sensitivity analyses are summarized in Figure 2 and in eTable 2 through eTable 13 in the Supplement. Overall, these sensitivity analyses yielded consistent findings, where current use of GLP-1 analogues was associated with an increased risk of bile duct and gallbladder disease, with HRs ranging between 1.59 and 1.84. In contrast, current use of DPP-4 inhibitors was not associated with an increased risk of bile duct and gallbladder disease, with HRs ranging between 0.97 and 1.02.

In this population-based cohort study, the use of DPP-4 inhibitors was not associated with an increased risk of bile duct and gallbladder disease. In contrast, the use of GLP-1 analogues was associated with an overall 79% increased risk of bile duct and gallbladder disease, which tended to be elevated in the first 180 days of use. Moreover, GLP-1 analogues were also associated with a 2-fold increased risk of cholecystectomy. Overall, these findings remained consistent in several sensitivity analyses.

To our knowledge, this is the first population-based study investigating the association between incretin-based drugs and the incidence of bile duct and gallbladder disease. Our findings are in accord with the results of a recent large randomized clinical trial investigating the effects of a high-dose GLP-1 analogue (liraglutide) in reducing body weight in overweight and obese patients.⁶ In this trial, liraglutide use was associated with substantial weight loss, but also an increased risk of gallbladder-related events (eg, cholelithiasis and cholecystitis) vs placebo (61 of 2481 and 12 of 1242, respectively). The number of cholecystectomies performed owing to cholelithiasis and cholecystitis was also imbalanced between the groups (liraglutide, 40 of 2481 vs placebo, 6 of 1242). In a similar trial in which the study population comprised overweight patients with type 2 diabetes, the number of patients with gallbladder-related events was also increased in the liraglutide group vs the placebo group (9 of 635 and 1 of 212, respectively).¹⁶

The rapid weight loss associated with GLP-1 analogues seems to be a logical mechanistic explanation for the increased risk of bile duct and gallbladder diseases. Weight loss leads to supersaturation of cholesterol in the bile, a known risk factor for gallstones.^17,18 The effect of rapid weight loss may be related to the fact that we did not find any significant association with longer GLP-1 analogue use (>180 days). However, in the liraglutide trials, the risk of gallbladder-related events was present across different weight-loss categories,¹⁹ suggesting an alternative mechanism. As was observed in a small crossover trial of exenatide,²⁰ use of GLP-1 analogues may result in reduced gallbladder emptying. Furthermore, experimental studies indicated that cholangiocytes are susceptible to GLP-1 and respond with increased proliferation and functional activity, which may increase the risk of duct obstruction.^3,4 Therefore, although the weight loss effect of GLP-1 analogues must be taken into account, it is possible that GLP-1 analogues increase the risk of gallbladder-related events via mechanisms independent of the weight loss effect, or by potentiating its effect.

The fact that the increased risk of bile duct and gallbladder disease was limited to GLP-1 analogues and not DPP-4 inhibitors could be explained by their different effects on GLP-1 pharmacologic factors. Although DPP-4 inhibitors increase the half-life of endogenous GLP-1, the GLP-1 analogues act as direct exogenous agonists to the GLP-1 receptor, leading to a stronger incretin action.^21,22 This direct mechanism explains the greater incidence of other gastrointestinal tract adverse effects (eg, nausea and vomiting) and the substantially greater weight loss observed with GLP-1 analogues compared with DPP-4 inhibitors.^2,23,24

Our study has many strengths. First, the use of a large cohort study of patients with type 2 diabetes allowed us to identify a significant number of bile duct or gallbladder events and to corroborate in a real-world setting a potential adverse event previously reported in randomized clinical trials. Second, the quality of the information recorded in the CPRD is high, and the use of the HES database to identify the outcome likely maximized the validity of case ascertainment. Finally, the use of a time-dependent exposure definition eliminated immortal time bias.²⁵

This study has some limitations. Since diabetes severity itself is a known risk factor for bile duct or gallbladder disease,²⁶ confounding by indication must be considered. To minimize this potential bias, we compared incretin-based drugs, which are second- to third-line therapies, with current use of 2 or more oral antidiabetic drugs. In addition, the models were adjusted for duration of treated diabetes, as well as important potentially confounding factors, including HbA_1c levels and diabetic macrovascular and microvascular complications as proxies of disease severity. With respect to drug exposures, we had to make the assumption that prescriptions written by general practitioners adequately reflect those of specialists (which are not recorded in the CPRD) and assume appropriate compliance by the patients who were prescribed these drugs. Although such potential misclassifications could have underestimated drug exposure, we do not expect differential renewal or compliance between the exposure groups. With respect to the outcome assessment, GLP-1 analogues are known to be associated with frequent gastrointestinal tract symptoms (including abdominal pain) and possibly pancreatitis and elevations of pancreatic enzyme. Such signs and symptoms may theoretically lead to overdetection, or perhaps misdiagnosis, of bile duct and gallbladder disease in patients taking GLP-1 analogues. However, while overdetection or misdiagnosis is possible, we believe that it is unlikely to have affected our findings. First, the most frequently gastrointestinal tract adverse reaction reported with GLP-1 analogues is mild to moderate nausea, a condition that is unlikely to necessitate hospital management. Second, even if patients presented with severe abdominal pain, which could be confused with biliary-type pain, a misdiagnosis of bile duct and gallbladder diseases (and a fortiori the diseases leading to cholecystectomy) would be highly unlikely, given that these conditions are typically confirmed with additional clinical signs and abnormalities on test results, as well as specific imaging findings. Third, our outcome definition was based on a primary admission diagnosis (ie, the diagnosis responsible for the admission), which we believe minimized potential misdiagnoses. Reassuringly, we obtained similar results in a sensitivity analysis that excluded and censored patients with pancreas-related conditions. Finally, some variables had missing information, such as body mass index, HbA_1c levels, and smoking status (range, 0%-18%). However, we obtained nearly identical results in a sensitivity analysis using multiple imputation methods.

In this large population-based study, we observed an increased risk of bile duct and gallbladder disease with the use of GLP-1 analogues while no association was observed for DPP-4 inhibitors. Although further studies are needed to confirm our findings and the mechanisms involved, physicians prescribing GLP-1 analogues should be aware of this association and carefully monitor patients for biliary tract complications.

Accepted for Publication: February 15, 2016.

Corresponding Author: Laurent Azoulay, PhD, Centre for Clinical Epidemiology, Lady Davis Institute, Jewish General Hospital, 3755 Cote St-Catherine, H425.1, Montreal, Quebec H3T 1E2, Canada (laurent.azoulay@mcgill.ca).

Published Online: August 1, 2016. doi:10.1001/jamainternmed.2016.1531.

Author Contributions: Dr Azoulay had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Faillie, Yu, Hillaire-Buys, Barkun, Azoulay.

Acquisition, analysis, or interpretation of data: Faillie, Yu, Yin, Azoulay.

Drafting of the manuscript: Faillie.

Critical revision of the manuscript for important intellectual content: Yu, Yin, Hillaire-Buys, Barkun, Azoulay.

Statistical analysis: Faillie, Yin, Barkun, Azoulay.

Obtained funding: Azoulay.

Study supervision: Azoulay.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was funded by the Canadian Institutes of Health Research.

Role of the Funder/Sponsor: The funding source had no influence on the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.