GLP-1 based agents and acute pancreatitis

BMJ 2013; 346 doi: http://dx.doi.org/10.1136/bmj.f1263 (Published 27 February 2013) Cite this as: BMJ 2013;346:f1263
  1. Edwin A M Gale, emeritus professor of diabetic medicine
  1. 1Department of Diabetes and Metabolism, Southmead Hospital, Bristol BS10 5NB, UK
  1. Edwin.Gale{at}bristol.ac.uk

Drug safety falls victim to the three monkey paradigm

Investment companies knew that the Food and Drug Administration safety database carried a signal for acute pancreatitis with the antidiabetic drug exenatide (a glucagon-like peptide 1 (GLP-1) agonist) in 2006, a year before the agency alerted doctors1—a curious reflection on the way we mix business with medicine. The signal had reached astronomical dimensions (more than 10 times that in control drugs) by 2011 and has accelerated since.2 Furthermore, all GLP-1 based agents that have been on the market for more than two years have also generated a signal for acute pancreatitis, suggesting a class effect.

The regulators asked companies to provide more data, and companies have responded with studies showing that acute pancreatitis is more common in diabetes than previously thought and that clear evidence of an increased risk of pancreatitis with GLP-1 based treatments is lacking.3 Warnings on the label notwithstanding, the industry has been able to maintain that the problem does not exist—and has a huge incentive to do so.

This is no longer tenable. A report in JAMA Internal Medicine describes a case-control study of more than a million people with diabetes, which yielded 1269 cases of acute pancreatitis in people aged 35-64 years using exenatide or sitagliptin, and an equal number of cases in people on non-GLP based drugs for diabetes. However, after multiple adjustments, current users and recent users (one month to two years) of GLP-1 based treatments had a twofold increased risk of acute pancreatitis (adjusted odds ratio 2.24 (95% confidence interval 1.36 to 3.68) for current use and 2.01 (1.27 to 3.18) for recent use) compared with those taking non-GLP based drugs for diabetes.4 A company sponsored study previously found an increase in episodes of pancreatitis in recent users of exenatide, but it discounted the observation because those affected were no longer taking the drug.3

Should we be worried about this? Very much so. GLP-1 is a pleiotropic agent that has many actions apart from its therapeutic effects in promoting insulin secretion, inhibiting glucagon release, delaying gastric emptying, and reducing appetite. It also interacts, for example, with receptors in the heart, kidneys, thyroid, and exocrine pancreas. Furthermore, GLP-1 is a very short acting peptide, and the consequences of long term pharmacological stimulation in humans are unknown. GLP-1 promotes cell replication in some tissues, and it was hoped that it would promote pancreatic β cell regeneration until this action was found to be restricted to immature rodents. This may have distracted attention from the fact that it also stimulates pancreatic duct cells to divide, regardless of age. It has been known for more than a decade that animal pancreas tissue increases in weight on exposure to GLP-1 and that this must represent overgrowth of the exocrine pancreas.5

Further observations in experimental animals prompted the hypothesis that overgrowth of pancreatic duct cells produces occasional obstruction of the smallest ducts, with the potential to cause subclinical pancreatic inflammation in many users and full blown acute pancreatitis in rare instances.6 Acute pancreatitis is unpleasant enough, but the major concern relates to subclinical inflammation of the pancreas. Postmortem analysis of humans exposed to GLP-1 based agents has yet to be reported, but it is well known that concentrations of pancreatic enzymes rise in animals and humans taking GLP-1 based drugs compared with other treatments for diabetes. Companies have modestly omitted these data from their published trials, sometimes with the comment that no “clinically significant” changes were seen. As a result, there is only one formal description of the phenomenon—from an independent group—in the literature.7 Subclinical increases in enzyme concentrations may not prove subclinical pancreatitis, but they provide no reassurance about its absence.

One reason why the merest possibility of pancreatitis has been contested so vigorously is that all forms of pancreatitis, clinical or subclinical, predispose to carcinoma of the pancreas. The harbingers of this unpleasant cancer, known as pancreatic intraepithelial lesions, are widely present as potential seeds of cancer in the adult population. These lesions carry the GLP-1 receptor, as do pancreatic carcinomas. Increased reporting of pancreatic cancer was independently noted in both the FDA and German regulatory databases.2 8

Why have the companies been so slow to respond to this threat? Because of the “three monkey paradigm,” which operates as follows. Companies are legally responsible for monitoring the safety of their own products, but self evidently cannot be held responsible for tackling a safety concern that does not exist. A concern that can be plausibly doubted or denied carries no legal liability, whereas one that gives rise to serious consideration (even in internal emails, which are discoverable) leaves the door wide open to litigation. Inviting companies to monitor the safety of their own products thus provides them with the strongest possible incentive for failing to do so, an instance of the law of unintended consequences. The three monkeys, who neither hear nor see nor speak, have been allowed to flourish at the heart of our system for protecting the public. The regulators should not follow this example.


Cite this as: BMJ 2013;346:f1263


  • Competing interests: I understood the BMJ Group policy on declaration of interests and declare the following interests: I have provided expert testimony in litigation concerning exenatide.

  • Provenance and peer review: Commissioned; not externally peer reviewed.