Every observational dataset published on GLP-1 receptor agonists in inflammatory bowel disease points the same direction — fewer steroids, fewer hospitalisations, fewer bowel resections. The single study built to remove the confounding found none of it. A meta-analysis of 11 studies and 16,242 IBD patients on GLP-1s reported odds of IBD-related surgery at 0.46 (95% CI 0.32–0.67) versus non-users, with a hazard ratio of 0.61 (PMID 41071055). Real-world abstracts presented at the January 2026 Crohn's & Colitis Congress went further, reporting intestinal resection at 6.5% versus 20.8% and mortality at 4.6% versus 22% in a matched ulcerative colitis cohort.
Then in June 2026 a target trial emulation ran the same question against OptumLabs claims data with the selection effects deliberately engineered out — stable patients only, no steroids, no recent hospitalisation or surgery, propensity matched one to one. One-year relapse risk on GLP-1s: 7.9%. Off them: 7.9%. Relative risk 1.01 (PMID 42309434). That is the paper the summaries circulating this week mostly leave out, and it is the one that changes how the rest should be read.
Research-context information only. Compounds discussed below are research peptides and supplements; some are investigational drugs not approved by the FDA. Semaglutide, tirzepatide and liraglutide are active ingredients in FDA-approved prescription products; the research-peptide and compounded forms discussed and linked below are not FDA-approved and are sold for research purposes only. None of the compounds named here is approved to treat inflammatory bowel disease. Protocols, doses, and reactions reported come from published research and self-reported community sources. Possession or use of investigational drugs outside an authorized clinical trial may be illegal in your jurisdiction. This article reports what has been documented, not what should be done. Consult a licensed physician for personal medical decisions.
The numbers driving the headlines
The pooled evidence base is the Bayoumy meta-analysis published in the Journal of Crohn's and Colitis in November 2025 — 11 studies, 16,242 IBD patients treated with GLP-1 receptor agonists, searched through March 2025.
| Endpoint | Reported effect | Detail |
|---|---|---|
| IBD-related surgery (hazard ratio) | logHR 0.61 (95% CI 0.44–0.84) | I² = 0% |
| IBD-related surgery (event frequency) | OR 0.46 (95% CI 0.32–0.67) | I² = 42% |
| Hospitalisation | Lower in BMI ≥ 30 subgroup only | Sensitivity analysis |
| Weight loss at 3 months — semaglutide | −9.6 kg (95% CI −12.0 to −7.2) | |
| Weight loss at 3 months — liraglutide | −9.4 kg (95% CI −13.0 to −5.8) | |
| Weight loss at 3 months — tirzepatide | −11.8 kg (95% CI −18.3 to −5.4) |
The surgery finding is the one with the tightest statistics behind it — zero heterogeneity on the hazard-ratio analysis. The hospitalisation finding did not hold across the whole population and only appeared once the analysis was restricted to patients with a BMI at or above 30, which is a useful hint about what is actually driving the association.
Two abstracts at the 2026 Crohn's & Colitis Congress in January produced the more dramatic figures. A Mayo Clinic group reported on 580 IBD patients treated with a GLP-1 — 322 with ulcerative colitis, 258 with Crohn's disease — matched against IBD controls not taking one. In the ulcerative colitis cohort the reported rates were corticosteroid use 46.6% versus 85.1%, hospitalisation 40.9% versus 65.9%, intestinal resection 6.5% versus 20.8%, and mortality 4.6% versus 22%.
A separate TriNetX analysis presented by Nakul Ganju's group at Howard University Hospital looked specifically at Crohn's disease: 546 GLP-1 users matched one to one against 546 non-users over 12 months. Steroid dependence came in at 52.9% versus 62.8%, hospitalisation at 2.3% versus 3.7%, hazard ratio 0.74, with no increase in major abdominal surgery. Those Crohn's numbers are a good deal more modest than the ulcerative colitis figures, which is itself worth noticing — several secondary write-ups of the congress have reprinted the ulcerative colitis percentages under a Crohn's heading.
The study that found nothing
The target trial emulation published in Clinical Gastroenterology and Hepatology in June 2026 is methodologically the strongest thing in this literature, and it is the only one that reports a null.
Its design is the point. A target trial emulation specifies the randomised trial you would run if you could, then constructs it from observational data with the same eligibility rules, the same assignment procedure and the same follow-up window. Here that meant OptumLabs Data Warehouse claims from 2018 to 2023, restricted to IBD patients who were genuinely stable at baseline — no steroid use, no IBD-related hospitalisation or surgery, and a stable dose of IBD therapy for more than six months. Two cohorts were built: one on 5-aminosalicylates or no IBD-directed medication, one on advanced therapies or immunomodulators.
In the first cohort, 2,028 GLP-1 initiators were matched one to one against non-initiators. Mean age 63, 63% female, 71% with ulcerative colitis. The one-year risk of relapse — a composite of IBD-related hospitalisation, surgery, or prednisone use — was 7.9% in both arms. Relative risk 1.01 (95% CI 0.82–1.24). Safety outcomes were 7.0% versus 7.9%, relative risk 0.88 (95% CI 0.71–1.10). Thirty-six percent of patients discontinued the GLP-1.
Nothing in that result contradicts the safety picture. What it contradicts is the efficacy picture — the idea that starting a GLP-1 does something to the disease.
Why the two designs disagree
The gap is not a data problem, it is a selection problem, and the authors of this literature say so themselves.
- Confounding by indication. A gastroenterologist starts a weight-loss drug in a patient whose colitis is quiet. A patient flaring toward a resection is not the patient being handed a new appetite-suppressing injectable. The comparison cohorts in the retrospective studies are therefore not comparable on the exact axis that predicts the outcome being measured.
- Healthy-adherer bias. People who fill, store, reconstitute and self-inject a weekly medication for twelve months are systematically more adherent to everything else, including their biologic. One of the congress findings was in fact better persistence with advanced IBD therapies in the GLP-1 group — which is a plausible mediator, but also exactly what a healthy-adherer effect looks like.
- Database overlap. A systematic review of 33 studies across immune-mediated inflammatory diseases, 20 of them in IBD, explicitly declined to perform a meta-analysis because of significant heterogeneity and overlapping data originating from the same insurance databases (PMID 41077375). Several of the "independent" cohorts in this field are drawing on the same underlying claims population.
- Obesity is doing work of its own. Obesity is an established predictor of worse IBD outcomes — more hospitalisation, more surgery, less steroid-free remission. A drug that removes 7–12 kg from an obese IBD patient has an entirely mechanical route to better outcomes that says nothing about anti-inflammatory activity. The meta-analysis's own sensitivity analysis, which found the hospitalisation benefit only at BMI ≥ 30, is consistent with that reading.
The one thing both camps agree on is safety. The Siranart systematic review pooled 10 observational studies and 10,362 IBD patients, 3,479 of them on GLP-1 receptor agonists, and found no increased risk of corticosteroid use (OR 0.93, 95% CI 0.17–5.16), treatment escalation (OR 0.70, 95% CI 0.06–7.62), or IBD-related surgery, alongside a 6.67% total body weight reduction and a mean absolute loss of 7.33 kg (PMID 41212197). Those confidence intervals are enormous — the surgery interval runs from 0.00 to 313.88 with I² of 92.5% — so the honest reading is "no signal detected," not "proven safe."
What it changes for anyone running a GLP-1 protocol
Nothing here alters legality, availability or pricing for any compound. What it changes is the expectation attached to the protocol.
The defensible version of the claim is narrow: in people who have IBD and obesity, GLP-1 receptor agonists produce roughly the same weight loss they produce in everyone else, and there is no detected signal of disease worsening. The TriNetX comparison is the cleanest evidence for that first half — 47,424 patients with IBD and obesity, 150 of them on semaglutide, mean total body weight change −16 ± 13.4 pounds versus −18 ± 12.7 pounds in matched patients without IBD, p = 0.24 (PMID 38642103). Having IBD did not blunt the weight response.
The indefensible version is that a GLP-1 is a colitis treatment. No randomised controlled trial has tested that, the one quasi-experimental design that approximated one found nothing, and the compounds involved are approved for diabetes, obesity and cardiovascular risk reduction — not for inflammatory bowel disease.
Current per-vendor $/mg, stock and COA coverage for the compounds named above — sold for research purposes, not as an IBD therapy — sit on Best Semaglutide Vendors, Best Tirzepatide Vendors and Best Liraglutide Vendors, with active discount codes across recommended vendors on the deals page.
