Can Ozempic Cause Gallstones? 27 Extra Cases per 10,000 a Year

Ozempic can cause gallstones, and the largest analysis of this drug class puts the increase in gallbladder disease at "an additional 27 cases per 10 000 persons treated per year."¹ Ozempic's own label reports gallstones in 1.5% of patients on 0.5 mg.² That absolute figure is the one to hold onto, because the relative risk quoted alongside it sounds far more alarming than the finding deserves.

Product labels complicate the picture further, in a useful direction. They disagree with each other about why it happens.

What did the 76-trial analysis find?

The 76-trial analysis found a relative risk of 1.37 for gallbladder or biliary disease, pooled across 103,371 participants.¹ Its confidence interval runs 1.23 to 1.52 and heterogeneity was zero, so the trials agreed with each other closely.¹

Broken into components, the pattern holds across every endpoint:

Only the last of those crosses 1.0, so biliary tract cancer did not reach significance.¹ Gallbladder removal carried the largest point estimate at 1.70. That matters more than the composite, because surgery is the outcome a person actually experiences.

How big is that in absolute terms?

In absolute terms the increase is 27 additional cases per 10,000 people treated per year, with a range of 17 to 38.¹ Its authors describe the increase in their own words as small: "the overall absolute risk increase was small (an additional 27 cases per 10 000 persons treated per year)."¹

Put the two framings side by side and the gap is the whole problem. A 37% increase in relative risk reads as a serious hazard. Twenty-seven extra cases per 10,000 per year reads as roughly one additional event for every 370 people treated for a year.

Arithmetic connecting them has a missing input. A relative increase of 37% converts into a count of extra events only against the baseline rate those trials started from, and the paper reports the 27 extra cases without reporting that baseline. Change the baseline and the same relative risk produces a different number of patients, so the two figures cannot be reconciled from what is published.

Both numbers describe the same finding. Neither is wrong. Relative risk tells you how much the rate moved, and the absolute figure tells you how much room there was to move. A rate that starts low stays low after a 37% increase.

What do the individual product labels report?

Each individual product label reports its own trial rates. Those rates are not the same. These four were read from the FDA label documents directly.

A warning applies to every comparison in that table. These rates come from separate trial programs with different populations, durations and reporting practices, so reading across the rows is not a head-to-head comparison and nothing here establishes that one product is safer than another.

Does Zepbound carry the same gallstone risk as Wegovy?

On the label figures, Zepbound's gallstone signal is close to absent while its cholecystitis signal is not. Cholelithiasis ran 1.1% on Zepbound against 1% on placebo, a difference of a tenth of a percentage point, while cholecystitis ran 0.7% against 0.2%.⁴ Cholecystectomy was reported in 0.2% of Zepbound patients and no placebo patients.⁴

Wegovy injection reports the opposite shape. Cholelithiasis ran 1.6% against 0.7% on placebo, more than double, while cholecystitis ran 0.6% against 0.2%.³

On their own labels, then, one product shows a gallstone difference and barely any inflammation difference beyond it, and the other shows an inflammation difference with almost no gallstone difference underneath. That is an odd pairing on its face, and neither label offers a mechanism that would produce the one without the other. Both report both outcomes and neither connects them. Our comparison of semaglutide and tirzepatide covers what else separates them, and the three-way version with retatrutide adds the newer molecule. Neither label was designed to support that contrast. Cross-trial figures are the weakest way to draw one.

Does the risk differ between weight loss and diabetes?

Substantially, and the gap between the two is the strongest subgroup finding in the analysis. Trials run for weight loss gave a relative risk of 2.29 across 13 trials and 11,282 patients, against 1.27 across 63 trials and 92,090 patients for diabetes and other indications, with an interaction P value below .001.¹

Nearly double the relative risk in the weight-loss population, from the same drug class. Weight-loss trials use higher doses and produce more weight loss, and both of those track with the risk separately in this same analysis.

Our guide to what happens after significant weight loss covers the broader version of this. Losing a lot of weight quickly is the context in which gallbladder events cluster, whatever is driving the loss.

Weight here rests on the interaction P value below .001. Interaction tests ask whether two subgroups differ from each other instead of whether each one differs from no effect, and a value that small says the weight-loss and diabetes populations are not drawing from the same risk.

Trial counts cut the other way on confidence. Thirteen weight-loss trials against 63 diabetes trials means the 2.29 estimate rests on about an eighth of the participants, and the authors flag possibly underpowered subgroups among their limitations.¹

Do the labels agree on why it happens?

Two labels do not agree on why it happens. Their disagreement sits on the central mechanistic question, between two documents both current.

The Zepbound label states plainly: "Acute gallbladder events were associated with weight reduction."⁴ Read on its own, that attributes the events to the weight loss and not to the drug.

Wegovy's label says the opposite. Gallbladder disease occurred more often with Wegovy than with placebo even after accounting for the degree of weight loss.³ If that adjustment holds, something other than the weight loss is contributing.

Neither label shows its working, so a reader cannot check either claim. Wegovy's statement is the stronger analytic claim, because a result that survives adjustment for the main confounder is harder to explain away than an unadjusted association. It also carries more obligation to show the adjustment. The label does not.

Does the dose matter?

Dose probably does matter, though the evidence points two ways, with one awkward exception on a label.

In the meta-analysis, high-dose trials gave a relative risk of 1.56 across 54 trials and 61,962 patients, while low-dose trials gave 0.99 across 33 trials and 16,952 patients, with an interaction P value of .006.¹ A relative risk of 0.99 is no effect at all, so on these figures the low-dose groups showed nothing.

Ozempic's label runs the other way. Cholelithiasis was reported in 1.5% of patients on 0.5 mg and 0.4% on 1 mg, so the higher dose carried the lower rate.² Those are pooled placebo-controlled trial figures with small event counts, not a dose-response experiment, and a label reports what the trials recorded without reconciling it.

One more wrinkle sits in the same place. No significant increase appeared for oral semaglutide at the doses used in diabetes trials, while the current Wegovy tablet label reports the highest adult cholelithiasis rate of the four products here, at 2.5% against 1%.¹³ Dose and indication changed between those two, so the two findings are not in direct conflict, and anyone citing the older null result for oral semaglutide is citing a different dose.

What does the paper say about dose in its own discussion?

Its own discussion contradicts its results table on dose, and anybody citing it should know that. The Results report significant interaction P values for both dose, at .006, and duration, at .03.¹ Its Discussion then describes those same associations as not significant: "Higher doses and longer duration of GLP-1 RAs treatment were also associated with increased risk of gallbladder or biliary diseases" and "although the association was not statistically significant."¹

One reading reconciles them. That same Discussion says higher risk was associated with higher doses "in weight loss trials" and that "there was a similar trend to increased risk with higher doses in the diabetes trials," so the not-significant remark may refer to the within-subgroup dose trends instead of to the interaction tests.¹

Nowhere does the paper say which it means. I read both sections and the ambiguity is in the source and not in the reading, so the safe citation is the interaction P values from the Results, with the Discussion sentence noted.

What happens when the comparator is another active drug?

Against another active drug the effect disappears. Against placebo the relative risk was 1.41 across 45 trials and 80,281 patients. Against an active comparator that was not a GLP-1 drug it was 1.03, with a confidence interval of 0.74 to 1.44 across 36 trials and 25,433 patients.¹ Those two did not differ significantly from each other, at an interaction P value of .08.¹

A relative risk of 1.03 against an active comparator is no detectable difference. That finding gets quoted less than the headline and it deserves more attention, because the practical question is rarely drug against nothing.

Treatment duration points the same way as dose. Trials longer than 26 weeks gave 1.40 across 53 trials and 90,417 patients. Trials of 26 weeks or less gave 0.79 across 24 trials and 13,401 patients, with an interaction P value of .03.¹ Short exposures showed nothing.

Which drugs showed no significant increase?

Four of the drugs analyzed showed no significant increase on these figures. Subcutaneous semaglutide came in at 1.28 with an interval of 0.99 to 1.65, just crossing 1.0, and exenatide at 1.23 with an interval of 1.00 to 1.52, touching it.¹ Oral semaglutide, lixisenatide and albiglutide showed no significant increase.¹

Two drugs did reach significance: liraglutide at 1.79, interval 1.45 to 2.25, and dulaglutide at 1.35, interval 1.06 to 1.73.¹ Subcutaneous semaglutide at doses of 1.0 mg or above reached 1.58, interval 1.13 to 2.22, even though the pooled semaglutide estimate did not.¹

That semaglutide split is instructive about how these estimates work. Pooling every dose produced an interval that just failed to clear 1.0, and restricting to 1.0 mg and above produced a clear signal, from the same drug in the same analysis. A null result at the pooled level is not evidence of no effect at the doses people actually take for weight loss.

Liraglutide's position at the top of the list is not explained anywhere in the analysis. That estimate is reported without an attributed reason, and duration and dose both tracked with risk separately in the same paper, so either could be behind it.¹

Tirzepatide appears nowhere in that list. Published in 2022, the analysis mentions tirzepatide in no sentence, so Zepbound's risk is not covered by the meta-analysis at all and the label figures above are the evidence for it.

Tirzepatide is also sold as Mounjaro for type 2 diabetes, and that label was not read for this piece. Given how far the weight-loss and diabetes estimates diverged in the analysis, at 2.29 against 1.27, the two tirzepatide products belong in separate questions until somebody reads both labels side by side.¹

Why is the pediatric rate higher?

Pediatric rates run higher on the Wegovy label, and the label does not explain why. Cholelithiasis was reported by 3.8% of Wegovy injection-treated patients aged 12 and up against 0% on placebo, and cholecystitis by 0.8% against 0%.³ Incidence of both was higher in pediatric patients than in adults, by the label's own account.³

Those adolescent figures are the highest on any of the four labels, and 3.8% is more than double the adult injection rate of 1.6%.³ No mechanism is offered on the label.

A placebo rate of 0% also deserves care. When the comparator arm records no events at all, a relative measure cannot be computed from those two numbers, and the apparent contrast depends partly on how few adolescents were enrolled. The label gives rates and not denominators in that passage.

Two obvious candidates sit unaddressed. Adolescents in those trials may have lost more weight, connecting back to the weight-reduction association, or the difference may reflect something about that population independent of the drug. The label reports the rates without separating those. Neither explanation is established.

How reliable is the underlying trial reporting?

The underlying trial reporting is weaker than the sample size suggests, and the authors say so. Biliary events were not a predefined safety outcome in most of the trials analyzed, and only 20 of the 76 trials clearly reported them.¹

A pooled analysis of 103,371 participants therefore rests on gallbladder data that most of the contributing trials were not set up to capture. The authors list that first among their limitations, alongside the absence of patient-level data and the possibility that small event numbers left subgroups underpowered.¹

Underreporting cuts in a specific direction. If events went unrecorded in trials that were not watching for them, the true rates in both arms are higher than reported, and what that does to the ratio between arms depends on whether the underreporting was even across arms. Nothing in the analysis establishes that it was.

One reason to doubt that it was even sits in the drugs themselves. Abdominal symptoms in a treatment arm have a competing explanation that a placebo arm does not, since gastrointestinal effects are expected on these drugs, and a trial not watching for biliary disease may file the complaint under the effects it was watching for. That would push the ratio down instead of up. It is the opposite direction from the usual worry about safety underreporting.

What should somebody do with this?

All three labels answer that, and they answer it the same way. Each one states that if gallstones or cholecystitis are suspected, gallbladder diagnostic studies and appropriate clinical follow-up are indicated.²³⁴

That is a clinical instruction and not a monitoring schedule, and none of the three labels sets out routine gallbladder screening for people taking these drugs. Our guide to the GLP-1 side effect data covers the symptoms that belong in that conversation.

What the labels do establish is that the drug makers and the agency both expect this to come up. A Warnings and Precautions section exists on all three products for acute gallbladder disease, a higher placement than an adverse reaction table entry, and all three carry the same instruction to investigate if it is suspected.²³⁴

The absolute figures are what make that proportionate. Twenty-seven extra cases per 10,000 treated years is a real signal and a small one. A labeled warning with no screening schedule attached is a reasonable regulatory response to a risk of that size.

What has not been measured?

Several things have not been measured, and the list is specific.

No head-to-head trial compares gallbladder outcomes between these products. Every cross-product comparison above is assembled from separate trial programs, the weakest form of comparison available. Other organ-level questions on this class have the same problem, and our guide to dehydration on retatrutide runs into it from the kidney side.

Whether the risk is a drug effect or a weight-loss effect is unresolved instead of unmeasured. Two current labels take opposite positions on it and neither shows the analysis behind its position. Our guide to why most peptides have no human evidence covers the general shape of a claim stated without its working.

Tirzepatide sits outside the meta-analysis entirely, so the only evidence here for Zepbound is its own label's trial pool. Our guide to the GLP-1 pills covers the oral forms, where the tablet rate is the highest adult figure in the table and the older analysis found nothing at diabetes doses.

The adolescent finding also has no mechanism attached. A rate of 3.8% against 0% is the largest gap on any of these labels, in the population where it is least explained.

Common questions

Can Ozempic cause gallstones?

How much do GLP-1 drugs increase gallbladder risk?

Is 27 cases per 10,000 a large increase?

Does Zepbound cause gallstones?

Which GLP-1 drug has the highest gallbladder risk?

Did any GLP-1 drug show no increased risk?

Is the gallbladder risk higher for weight loss than for diabetes?

Is it the drug or the weight loss causing gallstones?

Does a higher dose mean more gallbladder risk?

Does the risk go away compared with other diabetes drugs?

How long do you have to take it before the risk appears?

Why is the gallbladder rate higher in teenagers?

What is cholelithiasis on a GLP-1 label?

Do GLP-1 drugs cause gallbladder cancer?

Was tirzepatide included in the gallbladder meta-analysis?

How good is the trial data behind these numbers?

Should you be screened for gallstones on a GLP-1 drug?

Sources

1. He L, Wang J, Ping F, Yang N, Huang J, Li Y, Xu L, Li W, Zhang H. Association of Glucagon-Like Peptide-1 Receptor Agonist Use With Risk of Gallbladder and Biliary Diseases: A Systematic Review and Meta-analysis of Randomized Clinical Trials. JAMA Internal Medicine 2022;182(5):513-519. DOI 10.1001/jamainternmed.2022.0338. Read at source across two passes, Results and Discussion. The online supplement, holding the eTables, eFigures and per-subgroup absolute risks, was not read. Quoted for every relative risk and confidence interval above, all six outcome components, all five subgroup analyses with their trial and patient counts and interaction P values, the individual-drug estimates, the absolute risk sentence and the stated limitations. The conflict between the Results interaction P values and the Discussion's description of them is reported above as a conflict inside this one paper. ↩1 ↩2 ↩3 ↩4 ↩5 ↩6 ↩7 ↩8 ↩9 ↩10 ↩11 ↩12 ↩13 ↩14 ↩15 ↩16 ↩17 ↩18 ↩19 ↩20 ↩21 ↩22 ↩23 ↩24
2. Novo Nordisk. OZEMPIC (semaglutide) injection, for subcutaneous use. US prescribing information, revised 05/2026. Read at source from the FDA label document. Quoted for Warnings and Precautions 5.9 on acute gallbladder disease, the 1.5% and 0.4% cholelithiasis rates at 0.5 mg and 1 mg, the statement that cholelithiasis was not reported in placebo-treated patients, the follow-up instruction and the approved indications. Section 5.9 reports no cholecystitis rate, and that absence is cited above as an absence. ↩1 ↩2 ↩3 ↩4 ↩5 ↩6
3. Novo Nordisk. WEGOVY (semaglutide) injection, for subcutaneous use, and WEGOVY tablets. US prescribing information, revised 05/2026. Read at source from the FDA label document. Quoted for Warnings and Precautions 5.3 on acute gallbladder disease, the adult injection rates of 1.6% against 0.7% and 0.6% against 0.2%, the tablet rate of 2.5% against 1%, the pediatric rates of 3.8% against 0% and 0.8% against 0%, the statement that incidence was higher in pediatric than adult patients, and the statement that gallbladder disease occurred more often than placebo even after accounting for the degree of weight loss. That section reports no cholecystectomy rate and no tablet cholecystitis rate. ↩1 ↩2 ↩3 ↩4 ↩5 ↩6 ↩7 ↩8 ↩9 ↩10 ↩11
4. Eli Lilly. ZEPBOUND (tirzepatide) injection, for subcutaneous use. US prescribing information, revised 02/2026. Read at source from the FDA label document. Quoted for Warnings and Precautions 5.4 on acute gallbladder disease, the 1.1% against 1% cholelithiasis rates, the 0.7% against 0.2% cholecystitis rates, the 0.2% against none cholecystectomy rates, the sentence that acute gallbladder events were associated with weight reduction, the follow-up instruction and the approved indications. The section also notes similar cholelithiasis rates in the obstructive sleep apnea trials and does not mention biliary colic. ↩1 ↩2 ↩3 ↩4 ↩5 ↩6