Adults with chronic pancreatitis who were taking a GLP-1 drug were 41 percent less likely to start long-term opioid painkillers than closely matched patients who were not. That is the headline number from a retrospective study of 21,250 US patients published May 27 in Gastro Hep Advances ↗, and it is the first look at what these drugs do to the day-to-day burden of the disease rather than to blood sugar or weight.
Chronic pancreatitis is a slow, one-way destruction of the pancreas. The gland scars, stops making enzymes, and generates pain that is hard to treat. Much of the care is aimed at that pain: opioids first, then nerve blocks and endoscopic procedures, and in the worst cases surgery on the pancreas itself. Each step up that ladder is a marker of a disease getting harder to control.
The study followed patients down every rung of that ladder, and the GLP-1 group sat lower on all of them. New opioid prescriptions fell (hazard ratio 0.59, 95 percent confidence interval 0.52 to 0.67). Diagnoses of opioid use disorder dropped 42 percent (HR 0.58, 0.48 to 0.69). Patients were about half as likely to need a celiac plexus block, the injection that deadens the pancreatic nerves (HR 0.51, 0.34 to 0.78). They were 52 percent less likely to undergo an endoscopic pancreatic procedure (HR 0.48, 0.41 to 0.56), and 56 percent less likely to have major pancreatic surgery (HR 0.44, 0.33 to 0.59). Every one of those confidence intervals stays below 1.0, the line where a drug makes no difference. None of the five is the fluke of a single endpoint.
The numbers come from the TriNetX US Collaborative Network, a pool of hospital records. The authors took adults with chronic pancreatitis, split them by whether they had a GLP-1 receptor agonist on their medication list, then used 1:1 propensity score matching to build two groups of 10,625 that looked alike on age, sex, race, and major comorbidities before comparing three years of outcomes. The matched patients averaged 59.6 years old and were 53.7 percent women.
Here is the part to hold at arm's length. This is an association pulled from records, not a trial, and matching cannot balance the one thing that most separates these patients: the reason a doctor put them on a GLP-1 drug in the first place. People prescribed semaglutide ↗ or another GLP-1 receptor agonist ↗ tend to have diabetes or obesity, tend to be in regular care, and tend to be the kind of patient a health system tracks closely. Any of that could bend the outcomes without the drug touching the pancreas at all. The endpoints are also stand-ins for pain, not pain itself. Fewer opioid starts and fewer nerve blocks strongly suggest patients hurt less, but the database does not carry a pain score to prove it.
There is a real biological reason to run the study anyway. GLP-1 receptor agonists calm inflammation in animal models, and chronic pancreatitis is an inflammatory disease, so a class known for glucose and weight has a plausible route to the pancreas that has nothing to do with either. What the study cannot say is whether the drugs slowed the underlying scarring or just made a stretch of it more bearable.
The finding also sits against the class's own safety file. GLP-1 drugs carry a label warning for pancreatitis, and the acute-pancreatitis signal is one peptidemodel has covered before, most recently when a pharmacovigilance review found 308 pancreatitis flags for GLP-1 combinations that none held up ↗. A class flagged for possibly inflaming the pancreas now associates with fewer invasive interventions in people whose pancreas is already inflamed. Both things can be true, and untangling them is exactly the work a prospective trial with a real pain endpoint would have to do. Until one runs, this is a strong signal from hospital records, not a reason to prescribe.