ERCP threads an endoscope into the bile duct, and its best-known complication is pancreatitis — which makes a drug associated with fewer cases worth examining, given the long-running pancreatitis question covered in the pancreatitis evidence. This analysis reports exactly that, and then keeps going [1].
Acute pancreatitis came in at a risk ratio of 0.47 (95% CI 0.43–0.51). Cholangitis 0.56 (95% CI 0.50–0.62). Sepsis 0.51 (95% CI 0.46–0.57). Gastrointestinal bleeding 0.49. Biliary stricture 0.52. Repeat ERCP 0.53. Every major 30-day outcome the study measured was approximately halved.
The likeliest explanation is visible in how the procedure is used. ERCP is done urgently in people who are acutely unwell with obstructed, infected bile ducts, and electively in stable outpatients with stones or strictures. Someone on an ongoing GLP-1 prescription is disproportionately in the second group: well enough to be on maintenance medication, engaged with care, scheduled rather than admitted. Urgency of indication is not something propensity matching on recorded covariates can capture.
That is not a reason to dismiss the paper so much as to read it as a description of who is taking these drugs. The same signature — uniform improvement across outcomes with unrelated mechanisms — is what makes the mortality figures in the intracranial pressure meta-analysis implausible, and it is why the negative-control approach used in the FLOW subgroup analysis and elsewhere matters so much in claims research.
If there is a real effect here it is most likely on pancreatitis specifically, where a mechanism exists and where the concern started. Establishing it would need a design that separates elective from urgent procedures, or a randomized comparison — and 21,818 matched patients, however well balanced on paper, cannot substitute for either. The general difficulty of reading procedure outcomes in this literature runs through the thrombectomy cohort.