Incretin use was not associated with increased risk for acute pancreatitis

Patients who took incretin-based drugs had the same risk of acute pancreatitis as those who took other hypoglycemic medications or took neither type of drug, a case-control study found. Researchers looked at specific drug classes, current and past use of medication, and pancreatitis risk factors, finding no significant associations between acute pancreatitis with use of incretin-based medications for any subgroup.

The study was published in the June Diabetes Care. The following commentary by Joel Schectman, MD, was published in the ACP Journal Club section of the June 16, 2015, Annals of Internal Medicine.

Randomized controlled trials are not powered to find rare associations between treatments and adverse effects, and most cohorts are not large enough, nor sufficiently clinically detailed, to reliably evaluate such associations. Prior studies have yielded conflicting results about the association between incretin-based therapy and acute pancreatitis. The large, nested case–control study by Thomsen and colleagues did not find an association after adjustment for potential confounders.

The study has many strengths. First, it is truly population-based, with the entire Danish population of 5 million people. Although the cohort is large, it is not diverse—a caution for extrapolation. Second, it uses 10 (rather than the usual 1 to 2) matched controls per case, and therefore has more statistical power and is less subject to sampling error. Third, there were 12,868 first episodes of hospitalized acute pancreatitis (with good diagnostic validity) in this 8-year cohort—an impressively large number with which to model predictors. Fourth, with universal health care and prescription coverage, it is reasonable to assume that the prescription claims data are accurate.

As with any case–control study, the methodology and results of the study by Thomsen and colleagues may be questioned. Although the prescription data appear accurate, one wonders about ascertainment bias in other exposures. Only 3% of controls were recorded as obese—a remarkable tribute to the svelte Danish population or an example of the pitfalls of administrative databases. Because cases were, by definition, hospitalized at least once, a predominantly hospital-derived administrative dataset may have biased their methods toward greater capture of exposures among cases. To mitigate this risk, the authors might have used 2 smaller control groups, with 1 requiring an index hospital stay. Alternatively, it would be useful to know how frequently cases and controls were hospitalized and to adjust for this in the analysis.

Despite the questions raised, the study by Thomsen and colleagues showed similar small bivariable associations with pancreatitis among patients with ever use of all evaluated diabetes medication classes. Adjustment for potential confounders eliminated these associations, supporting the view that an association of diabetes with pancreatitis may be mediated by biliary comorbidities and/or obesity. A putative association of incretin-based therapies with acute pancreatitis has been further undermined.