Increased Risk of Acute Pancreatitis and Biliary Disease Observed in Patients With Type 2 Diabetes: A retrospective cohort study

A retrospective claims database analysis was performed using a proprietary research database containing eligibility information and pharmacy and medical claims data from a large commercial U.S. health plan providing coverage for physician, hospital, and prescription-drug services. The plan subscribers represent a geographically diverse sampling from all regions of the U.S., with the greatest proportions of members in the Midwest and South. The database includes medical and prescription drug benefit claims data for ∼14 million patients during 2007. Data derived from this source have been used for a variety of utilization, safety, and economic analyses (12,–14).

Eligible patients were ≥18 years and enrolled for at least 12 continuous months from 1999 to 2005, with at least 30 days of follow-up from the end of the 1-year enrollment (n = 9,249,211). The index date was defined as the date when the patient accrued 1 year of prior continuous enrollment (the baseline period).

Table 1 presents the patient selection process. Patients were assigned to the nondiabetic cohort if during the study period, they had no medical claims for diabetes (ICD-9 code 250.xx), no claim for an antidiabetic medication, and at least one medical claim processed (n = 6,947,299). Patients were assigned to the type 2 diabetic cohort if during the study period, they had a claim for type 2 diabetes (250.x0 or 250.x2) and a filled prescription for an antidiabetes medication (n = 386,369). Antidiabetes medications were defined as biguanides, sulfonylureas, thiazolidinediones, α-glucosidase inhibitors, meglitinides, d-phenylalanine derivatives, pramlintide, exenatide, or insulins. Patients receiving insulin-only therapy with a claim for type 1 diabetes (250.x1 or 250.x3) were excluded from the study (n = 352,633 patients remaining).

Only patients of known sex were included in either cohort. Cohorts were pair-matched 1:1 on sex and age (n = 352,569). The pairs with available claims data after matching were used for this analysis (n = 337,067).

Patients with acute pancreatitis (ICD-9 code 577.0), cholelithiasis (574.2x, 574.5x, 574.9x, 576.1x, and 576.2x), acute cholecystitis (574.0x, 574.3x, 574.6x, 574.8x, 575.0, and 575.12), and cholecystectomy (51.2x) were identified based on the presence of the relevant diagnosis code on any medical claim. To ensure that this was a study of disease incidence, patients were excluded if they experienced outcome events during the 1-year baseline period. Specifically, for the pancreatitis analysis, patients were excluded if they had any claim for acute or chronic pancreatitis during the 1-year pre-enrollment period. For the biliary analysis, patients were excluded if they had any claim for cholelithiasis, acute cholecystitis, or cholecystectomy during the 1 year pre-enrollment. Laboratory, radiology, and pathology claims were not used to identify outcome events.

The total time at risk began on the index date and ended on the date of the first outcome event, disenrollment date, or the end of study date—whichever occurred first. For each cohort, exposure-adjusted incidence rates (per 100,000 patient-years of exposure) and 95% CI were calculated using Wald's method (15) for the overall cohort, stratified by age category and sex. Exposure-adjusted incidence rate ratios (IRRs) and accompanying 95% CIs between the two groups and within each stratum were also calculated (16). Patients with a diagnostic code for acute pancreatitis, cholelithiasis, acute cholecystitis, or cholecystectomy served as the numerators for the incidence calculations. All patient-years within each cohort served as the denominators.

The incidence rate for pancreatitis in the type 2 diabetic cohort was 422 cases per 100,000 patient-years compared with 149 cases per 100,000 patient-years in the nondiabetic cohort. In the diabetic cohort, the incidence rate was relatively constant across age-groups in contrast to the nondiabetic cohort, in which age was positively correlated with the incidence of pancreatitis (Table 2).

Overall, the type 2 diabetic cohort was at 2.83-fold (95% CI 2.61–3.06) greater risk of pancreatitis compared with the nondiabetic cohort. Relative to patients of corresponding age without diabetes, the youngest type 2 diabetes age-group (18–30 years) had the highest IRR of acute pancreatitis (7.75 [95% CI 3.89–15.43]), whereas patients aged ≥65 years with type 2 diabetes had the lowest IRR (1.64 [1.36–1.98]). Because type 2 diabetes typically occurs in patients aged ≥45 years, the IRRs of pancreatitis were calculated among individuals aged older and younger than 45 years. The results indicated that type 2 diabetic patients between the ages of 18 and 44 years experienced a 5.26-fold (95% CI 4.31–6.42) increased incidence of pancreatitis and those patients aged ≥45 years had a 2.44-fold (2.23–2.66) increased incidence of pancreatitis compared with their nondiabetic counterparts of the corresponding age-group. For acute pancreatitis, the IRRs were similar between males and females in the two cohorts.

The incidence rate for biliary disease in the type 2 diabetic cohort was 1,411 cases per 100,000 patient-years compared with 737 cases per 100,000 patient-years in the nondiabetic cohort. In the diabetic cohort, the incidence rate was highest in the youngest (18–30 years) and oldest (≥65 years) age-groups, in contrast to the nondiabetic cohort in which age was positively correlated with the incidence of pancreatitis (Table 3). In both cohorts, the incidence of biliary disease was notably higher in women than in men, although the IRRs were similar between the sexes for the two cohorts.

Overall, the type 2 diabetic cohort had a 1.91-fold (95% CI 1.84–1.99) greater risk of biliary disease than the nondiabetic cohort. Relative to patients of corresponding age without diabetes, the youngest type 2 diabetes age-group (18–30 years) had the highest IRR of biliary disease (3.77 [95% CI 2.92–4.87]), whereas patients aged ≥65 years with type 2 diabetes had the lowest IRR (1.50 [1.37–1.65]).

Examination of the biliary disease subgroups revealed that cholelithiasis contributed ∼50% of the total incident cases of cholecystitis, cholecystectomy, and cholelithiasis among both cohorts, and the incidence of cholelithiasis in type 2 diabetic patients was considerably higher (1,229 cases per 100,000 patient-years) than that in patients without diabetes (647 cases per 100,000 patient-years) (data not shown).

This study suggests that patients with type 2 diabetes have an almost threefold greater risk of pancreatitis and twofold greater risk of biliary disease than patients without diabetes. The high risk of pancreatitis among younger patients with type 2 diabetes relative to their nondiabetic peers is particularly notable, although the clinical meaning of this finding needs to be elucidated.

This study was limited by the data available in a managed care claims database, given that claims data are collected for payment and not for research. Patient compliance with prescription medications and use of physician samples were not captured. In addition, data regarding other possible risk factors for pancreatitis (such as alcohol use, obesity, weight loss, and concomitant medications) were not available. Another potential limitation is error in disease ascertainment, given that diagnostic codes may be incorrectly coded or included as rule-out criteria rather than actual disease. For example, although we used a conservative algorithm for identifying patients with type 2 diabetes (ICD-9 code AND use of an antidiabetic agent), some patients with type 1 diabetes were probably included in the cohorts, particularly in the younger age-groups. It is also noteworthy that the incidence rate of pancreatitis for the nondiabetic cohort reported in this epidemiologic study is approximately threefold greater than published estimates for the general population (1,4). The higher incidence of pancreatitis found in this study may be representative of an increase in pancreatitis, claims miscoding for pancreatitis, population differences, or the reporting method. A study of the accuracy of ICD-9 codes for pancreatitis conducted in a large VA population reported excellent sensitivity (93%) but lower specificity (79%) for acute pancreatitis (17); hence, it is likely that false-positive reports of pancreatitis were included in this study. Although the reason for the higher incidence of pancreatitis observed in this study is unknown, it is likely to be nondifferential across the two cohorts such that the increased risk of pancreatitis observed among the patients with type 2 diabetes is valid.

Strengths of this study include the large sample size, which is necessary given that pancreatitis is a rare event. These data also allow for the examination of health outcomes in a “real world” setting including a nationwide sample of patients with diverse medical histories. Nonetheless, the data used for this study come from a managed care population, and results are applicable primarily to the prevalence of outcomes in managed care settings. Age and sex bias were controlled for by pair matching. Finally, these results are probably conservative, given that subjects with undiagnosed diabetes may have been included in the nondiabetic cohort, a problem that is not unique to claims data.

In summary, the nearly threefold increased risk of pancreatitis for patients with type 2 diabetes reported here, combined with the increasing prevalence of diabetes and the associated risk factors, may be contributing to a meaningful increase in the incidence of acute pancreatitis in the U.S. Further studies are required to confirm these findings and to identify causal factors that may account for the observed increased risk of pancreatitis associated with diabetes.

This study was supported by Amylin Pharmaceuticals and Eli Lilly. Dawn Nicewarner, a medical writer, an employee, and stockholder of Amylin Pharmaceutical, assisted in the preparation of this manuscript. No other potential conflicts of interest relevant to this article were reported.

We thank i3 Innovous for statistical analysis and valuable discussions.