Two systematic reviews published within a year of each other looked for bowel obstruction on this drug class, and both singled out liraglutide. One found it the only agent with a significantly raised risk, at an odds ratio of 3.00 (95% CI, 2.03 to 4.45).[1] The other found it the only agent with a significantly lowered risk, at an odds ratio of 0.44 (95% CI, 0.24 to 0.81).[2] Same molecule, same outcome, opposite signs, neither interval crossing one. That is not a rounding disagreement, and it is the most honest summary available of how firmly this particular risk is established.
What the labels actually added, and when
Ileus and intestinal obstruction appear on every US label in this class, and they appear in the same place on each: section 6.2, postmarketing experience, in a list that the labels themselves preface by saying the reports come from a population of uncertain size and that frequency cannot be reliably estimated. The semaglutide labels group them as “ileus, intestinal obstruction, severe constipation including fecal impaction”; the tirzepatide labels use the same wording.[3][4][5] No incidence is printed anywhere for either event, on any of the six labels.
What did get added in the last year is a different section. A Warnings and Precautions heading called Severe Gastrointestinal Adverse Reactions was dated 10/2025 on the semaglutide and liraglutide labels, 12/2025 on the tirzepatide diabetes label and 02/2026 on the tirzepatide weight label.[3] [4][5][6] That section names neither ileus nor obstruction. It carries a rate for severe gastrointestinal reactions as a category — liraglutide 4.8% against 1.4% on placebo, oral semaglutide 0.6% at 7 mg and 2% at 14 mg against 0.3%, injected semaglutide for diabetes 0.4% and 0.8% against 0% — and a single exclusion, that the drug is not recommended in severe gastroparesis.[5][6] [7] The condition that exclusion covers, and the emptying literature behind it, is a separate article.
The pharmacovigilance signal, and what it is counting
The number that put obstruction on the map came from VigiBase. Across 501,244 adverse reaction reports involving diabetes drugs, 452 intestinal obstructions involved an incretin-based drug, and they were reported more than four and a half times as often as with other diabetes drugs — reporting odds ratio 4.52 (95% CI, 3.87 to 5.28).[8]
The breakdown inside that paper is the part that should stop anyone from quoting the headline. The signal was larger for DPP-4 inhibitors at 8.66 (95% CI, 7.27 to 10.32) than for GLP-1 analogs at 3.05 (95% CI, 2.54 to 3.66).[8] DPP-4 inhibitors do not slow gastric emptying, do not produce the gastrointestinal effects this class is known for, and are not the drugs anyone worries about. The mechanism story predicts the opposite ranking from the one the database produced, which is what a reporting artifact looks like from the inside.
A 2026 comparative analysis of European reports adds the other failure mode. Intestinal obstruction accounted for 1.55% of tirzepatide reports even though it is not listed in that product’s European summary of characteristics — a finding the authors read as a possible signal and as a reminder that reporting patterns follow what readers have been told to look for.[9] Neither instrument has a denominator. Both count reports, not patients, and a report is filed by someone who already suspected the drug.
The cohort that had a denominator found nothing
Nationwide registers in Sweden, Denmark and Norway supplied the test. Among 121,254 new users of GLP-1 receptor agonists and 185,027 new users of SGLT2 inhibitors as the active comparator, 557 intestinal obstruction events occurred. The adjusted incidence was 1.3 against 1.6 per 1,000 person-years, a hazard ratio of 0.83 (95% CI, 0.69 to 1.01). In the companion cohort, 190,321 DPP-4 inhibitor users produced a hazard ratio of 1.13 (95% CI, 0.96 to 1.34).[10] The class with the eightfold reporting signal and the class with the threefold reporting signal both landed on no measurable excess once every treated person was counted, not only the ones who generated a report.
The claims study everyone cites, read to the end
A 2023 analysis in JAMA is the source of nearly every consumer claim on this subject. It drew 613 semaglutide users, 4,144 liraglutide users and 654 people on bupropion-naltrexone as the comparator, all coded for obesity and none for diabetes, and reported an adjusted hazard ratio for bowel obstruction of 4.22 (95% CI, 1.02 to 17.40).[11]
Three things inside that paper are rarely carried with the ratio. The incidence rates per 1,000 person-years were 0 for semaglutide, 8.1 for liraglutide and 1.7 for the comparator: the semaglutide arm recorded no bowel obstructions at all, so the class estimate is carried entirely by 73 liraglutide events against two.[11] The interval runs from 1.02 to 17.40, which is what a handful of events looks like when it is expressed as a ratio. And the paper’s own sensitivity analysis excluding hyperlipidemia moved bowel obstruction to 3.63 (95% CI, 0.87 to 15.10) — no longer significant.[11] A figure that survives in the abstract and dies in the sensitivity table is not a foundation for a purchase decision, and the same dataset’s gastroparesis result is handled in the motility article.
The randomized record is close to empty, and that is the finding
Three large syntheses of randomized data have now looked. A network meta-analysis of 50 trials and 192,359 participants prespecified incident small or large bowel obstruction as its primary outcome and found exactly one elevated agent, and it was not a GLP-1 drug: canagliflozin at an odds ratio of 2.56 (95% CI, 1.01 to 6.49), with the 300 mg dose at 3.42. Liraglutide came out protective at 0.44, an absolute risk reduction of 0.34% and a number needed to treat of 295.[2] A synthesis of 55 placebo-controlled trials in 106,395 participants prespecified paralytic ileus and intestinal obstruction alongside a dozen other outcomes, found increased risk only for cholelithiasis and reflux, and concluded that the class probably has little or no effect on the rest.[12]
The pooled observational picture is no firmer. Fourteen studies were screened and six meta-analyzed, covering 550,426 participants: odds ratio 1.95 with a confidence interval from 0.43 to 8.79 and heterogeneity of I² = 94%.[1] An interval that wide, around a point estimate near two, with almost all of the variance between studies rather than within them, is a way of saying the literature has not converged. The liraglutide subgroup inside it was the one clean compartment, at I² = 0%.
Routine practice gives the same answer from a fourth direction. A new-user, active-comparator study propensity-matched 65,238 pairs for semaglutide against dulaglutide, 20,893 pairs for tirzepatide against dulaglutide and 46,620 pairs for tirzepatide against semaglutide, with bowel obstruction as one component of the primary composite. The hazard ratios were 0.96, 0.96 and 1.07, every interval crossing one.[13] Whatever is true here is not a difference between the two molecules most people are choosing between.
Who is actually at risk
Obstruction is an anatomy problem before it is a drug problem. Adhesions — scar tissue described in the international guideline as the footprints of previous abdominal surgery — are the dominant cause, and the guideline is built entirely around patients who have had an operation.[14] A systematic review of the opposite group, 855 adults obstructed with no prior abdominal surgery at all, found adhesions still responsible for 36% (95% CI, 22% to 49%) and malignancy for 12%, with 74% managed without an operation.[15] The people for whom a slowed transit could plausibly tip into obstruction are the people who already have something narrowing the lumen.
The published case reports read that way. A 61-year-old woman on tirzepatide for over a year developed progressive constipation, then a closed-loop obstruction caused by dense adhesive disease and an internal hernia, requiring resection of 25 cm of necrotic small bowel. The authors name the adhesions as the direct cause and the drug as a possible contributor through worsening baseline constipation.[16] That is the realistic pathway: not a drug that obstructs a normal bowel, but a drug that hardens the stool in someone whose bowel already has a bottleneck. The stool side of that is measured in the constipation article, and prior abdominal surgery is also the setting covered in the post-bariatric article.
What a compounded vial changes
Every figure above comes from randomized trials, registries or claims records of FDA-approved product at labeled doses. Most sellers covered here dispense compounded semaglutide or tirzepatide, which is not FDA-approved and is not reviewed by the FDA for safety, efficacy or quality before it is dispensed, and no surveillance system separates compounded products out by outcome. A pharmacovigilance analysis of 81,078 GLP-1 reports, 707 of them compounded, found higher reporting odds of abdominal pain at 2.84, diarrhea at 1.59 and hospitalization at 2.35; obstruction is not among the reactions it reports as elevated.[17] A survey of 33 advertised compounded products found 48% combined the drug with cyanocobalamin, glycine, niacinamide, docusate or ondansetron, and that only 18% published a beyond-use date with storage conditions.[18] Docusate is a stool softener with no evidence base for subcutaneous use, which is a strange thing to find in a vial sold to people worried about their bowel.
Where the line sits
On the published evidence, ileus and bowel obstruction are labeled postmarketing events with no established incidence, no consistent randomized signal, and a disproportionality signature that ranks a class with no motility effect above the one that has it. The honest statement is that this is rare enough that four large instruments cannot agree on whether it is elevated at all — which is a different and weaker claim than either “proven safe” or the ratios circulating online. The wider tolerability picture is in what the trials recorded, and the dose schedule those decisions attach to is in the titration article.
The presentation that warrants an emergency department rather than a message to a prescriber is specific and does not resemble ordinary nausea: abdominal pain that is severe, crampy and escalating, a swollen and tight abdomen, vomiting that will not stop, and no passage of stool or gas. Anyone with prior abdominal surgery, a hernia or known adhesions should have that history on the record before a first dose, and whether an intake asks for it is noted in each provider review.