Of the 3,460 biliary side-effect reports filed to the FDA for the five major GLP-1 drugs, dulaglutide accounted for the cholangitis signal that stood furthest from semaglutide. Its proportional reporting ratio for cholangitis (bile-duct infection) was 1.65, the highest of any drug-by-outcome pair in the analysis. The same drug had a lower-than-baseline rate of biliary colic. Exenatide, the oldest agent in the class, had the highest cholelithiasis (gallstone) and cholecystitis (gallbladder inflammation) signals but the lowest bile duct stone signal. The class effect, in other words, did not hold.
A team published the disproportionality analysis online May 28 in Digestive Diseases and Sciences ↗. They pulled five biliary outcomes (cholelithiasis, cholecystitis, biliary colic, bile duct stone, cholangitis) from the FDA Adverse Event Reporting System for five drugs: semaglutide ↗ (Ozempic, Wegovy, Rybelsus), tirzepatide ↗ (Mounjaro, Zepbound), liraglutide ↗ (Victoza, Saxenda), exenatide ↗ (Byetta, Bydureon), and dulaglutide ↗ (Trulicity). They used semaglutide as the within-class reference and calculated proportional reporting ratios for each non-reference drug against it.
The total report counts give an idea of the class market share: semaglutide led at 1,797 reports, tirzepatide 1,363, liraglutide 1,033, exenatide 999, dulaglutide 574. After deduplication, the numbers split out into four broadly different drug fingerprints, plus the semaglutide baseline.
What each drug looked like
Compared with semaglutide, exenatide had the strongest cholelithiasis and cholecystitis signals (PRR 1.33 and 1.12) but the weakest signals for the two downstream outcomes: bile duct stones (PRR 0.39) and biliary colic (PRR 0.30). Liraglutide tracked exenatide on the gallbladder outcomes (PRR 1.21 and 1.07) but without the downstream pattern. Tirzepatide showed mildly elevated cholelithiasis and cholecystitis (PRR 1.15 and 1.05) plus reduced bile duct stones (PRR 0.58). Dulaglutide, the smallest reporter, was the outlier on cholangitis (PRR 1.65) and on biliary colic in the other direction (PRR 0.50).
Sensitivity analyses were broadly concordant. Subgroup analyses found the within-class heterogeneity to be sharpest on bile duct stones and biliary colic, which suggests the differences are not just reporting noise.
Why this is the news
The standard summary of GLP-1 biliary risk, in the prescribing information of every drug in the class and most clinical reviews, is that the class as a whole increases gallbladder events relative to placebo. That statement has held up for years in randomized trials. What it does not do is distinguish among agents. This paper takes the same regulatory database that built the original class warning and asks whether each agent contributes equally to it. The answer is no.
Two patterns are clinically interesting. First, the older agents (exenatide, liraglutide) lead the cholelithiasis and cholecystitis columns, while the newer agents (semaglutide, tirzepatide) sit closer to the middle of the within-class distribution. Whether that reflects molecule-specific pharmacology, weight-loss trajectory differences, or systematic differences in how the older drugs were prescribed and reported is not something a disproportionality analysis can answer. Second, dulaglutide stands alone on cholangitis. The mechanism (if any) is not obvious from the abstract; the authors flag it as a heterogeneity worth following up.
The clinical reading is narrow but real. A patient with prior gallstone disease, or with biliary anatomy concerns, has a slightly different risk profile depending on which GLP-1 the prescriber reaches for. The differences are subtle, the database is biased toward reported cases, and none of this should change first-line prescribing. But the assumption that all five drugs carry an identical biliary side-effect profile is not what the data say.
What FAERS analyses can and cannot do
FAERS is a passive pharmacovigilance database. It collects spontaneous reports from clinicians, manufacturers, and patients. Disproportionality analysis asks whether one drug-event pair is reported more often than the average rate of that event across the rest of the database. A high proportional reporting ratio is a hypothesis: this drug-event pair is over-represented relative to background. It is not an incidence rate, not a relative risk, and not a confirmed causal signal.
The standard caveats apply. Drugs marketed for longer (exenatide since 2005, liraglutide since 2010) have had more time to accumulate reports, and reporting rates change with media attention. Indication channeling matters: dulaglutide was disproportionately prescribed to older T2D patients with more comorbidity. The FAERS denominator is not patient-exposure but report-volume. None of these confound out the within-class differences, but they shape how the signal should be read.
Where this connects on peptidemodel
All five agents have cards on the platform: semaglutide ↗, tirzepatide ↗, liraglutide ↗, exenatide ↗, and dulaglutide ↗. The shared receptor target is GLP-1R ↗. Tirzepatide additionally engages GIPR, which may explain part of its position closer to semaglutide than to the pure GLP-1R agonists on these particular outcomes. The mechanistic question worth following up is whether the within-class heterogeneity tracks the molecular differences (structure, half-life, dual-receptor engagement, formulation) or some confound the database cannot remove.