This site has already reported the randomized evidence on gallstones, and why the relative and absolute versions of it feel so different. Here is the same question asked a completely different way, and the answer that came back is worth more than either estimate on its own.
What the cohort did
From a research network of 156,376 adults with type 2 diabetes, 43,077 GLP-1 users were matched one to one against controls, leaving 39,140 in each arm, and followed for gallstone disease, cholecystitis, pancreatitis, ERCP and gallbladder removal. [1]
At two years, gallstones came in at an adjusted odds ratio of 1.44, 95% CI 1.24 to 1.65. By three years it was 1.43, 95% CI 1.24 to 1.63, with cholecystitis at 1.45, 95% CI 1.14 to 1.83. Pancreatitis and ERCP showed no significant difference at any point.
The endpoint that arrived late
Gallbladder removal was not significantly raised at two years. By three years it was, at an adjusted odds ratio of 1.54, 95% CI 1.17 to 2.02.
That sequence makes sense clinically — stones form, then some of them cause trouble, then some of that trouble ends in an operating theater — and it has a consequence nobody prices. The outcome with a surgical bill attached is the one that takes longest to appear, so any study short enough to be affordable is systematically likely to miss it. A two-year analysis of this exact cohort would have reported gallstones and no excess surgery.
For somebody weighing the cost of a year on one of these drugs, that is the relevant shape: the cheap consequence shows up inside the window you are thinking about, and the expensive one shows up after — which is why the real total is harder to compute than a monthly figure suggests.
The molecule split, and why not to trust it
Semaglutide and dulaglutide were associated with higher gallstone rates. Liraglutide and exenatide were not.
That will be read as evidence the older drugs are gentler on the gallbladder. It is not. Far fewer people in a recent cohort take liraglutide or exenatide, and a subgroup with fewer users produces wider intervals and fewer significant findings whether or not any difference exists. A null in a small subgroup is a statement about the subgroup’s size — the same trap a within-class comparison sets when the arms are uneven.
What to do with it
Nothing dramatic. Gallstones on these drugs are an established, modest, well-replicated risk, and the absolute numbers remain small even though the relative ones sound large.
The practical points are two. Sudden severe pain in the upper right abdomen after a meal is worth taking to an emergency department rather than a support inbox. And if you are pricing a multi-year course, the gallbladder is one of the few complications with a known, quantified, delayed cost attached — which puts it in a different category from most of what gets listed in a price comparison.