A 3,478-patient match published this month in Crohn's & Colitis 360 puts GLP-1 receptor agonists ahead of bariatric surgery on every inflammatory bowel disease outcome the analysis tracked, even though the GLP-1 arm ended the study heavier.
That last clause is the story. For two decades, the framing in obesity with IBD has been that surgery is the structural solution, because it produces the largest and most durable weight loss, and weight loss is what reverses obesity-driven inflammation. The new analysis says that framing is upside down. Patients who never lost as much weight still had cleaner guts.
What the analysis did
The authors pulled records on every adult with IBD in the TriNetX network, which aggregates de-identified electronic health record data from more than 200 US health systems. They split the cohort into two groups, both with obesity: 12,965 patients on a GLP-1 receptor agonist ↗ who never had bariatric surgery, and 4,177 patients who had bariatric surgery but no GLP-1 RA. The drug arm covered every approved injectable in the class: dulaglutide ↗, semaglutide ↗, liraglutide ↗, exenatide ↗, lixisenatide, albiglutide, and the dual GIP/GLP-1 agonist tirzepatide ↗.
After propensity-score matching on age, comorbidities, baseline IBD therapy, and demographics, 3,478 patients remained in each arm. The drug arm was still older, sicker on paper, and a kilogram heavier per square meter than the surgery arm at the end of follow-up. Post-intervention body mass index was 34.5 versus 33.4. If weight loss were the variable doing the work, surgery should have won.
The outcomes
It did not. Across nine outcomes the authors prespecified, GLP-1 RA recipients had lower odds on every one, all with p-values below 0.0001.
Colectomy, the surgical endpoint nobody wants, was the most extreme. Patients on GLP-1 drugs were about one-eighth as likely to lose part of their colon (odds ratio 0.13, 95 percent confidence interval 0.09 to 0.20). Intestinal obstruction came in at 0.21, C. difficile infection at 0.22, lower-extremity deep vein thrombosis at 0.24, and pulmonary embolism at 0.27. The headline IBD outcome, disease flares, was 0.44, meaning a 56 percent lower odds of a flare. Initiation of an advanced IBD therapy (biologics or small-molecule advanced agents) was 0.45.
For a reader who is not used to odds ratios, the rough translation is that across a wide range of complications, the drug arm had between one-fifth and one-half the rate of the surgery arm. The exact numbers depend on the outcome. The direction does not.
Why the BMI gap matters
Most prior work in obesity with IBD has assumed that the path to better gut outcomes runs through weight. Bariatric surgery does deliver more weight loss than any GLP-1 drug, even tirzepatide ↗. So the conventional read of any head-to-head should have favored surgery on the inflammatory endpoints downstream.
The new data inverts that. It is the cleanest signal yet that GLP-1 receptor signaling is doing something to IBD physiology beyond pulling weight off. GLP-1 receptors are expressed on intestinal epithelium, on enteric neurons, and on multiple immune cell populations in the gut wall. Preclinical work has documented direct GLP-1-mediated effects on mucosal barrier integrity and on T-cell activation in colitis models. The TriNetX analysis cannot prove the mechanism, but it is consistent with one in which the drug works on the gut, not only on the scale.
The lower colectomy and flare rates also drag down the cascade. Fewer flares means fewer corticosteroid courses, fewer hospitalizations, fewer bowel resections, fewer thromboembolic events that come from inflamed bowel and steroid exposure together. That cascade probably explains why the deep vein thrombosis and pulmonary embolism rates fall so far in the GLP-1 arm. Less inflammation upstream, fewer clots downstream.
What it does not show
This is a propensity-matched retrospective cohort, not a randomized trial. The two arms are still not identical even after matching: the GLP-1 group had shorter follow-up, more diabetes, and more chronic kidney disease at baseline, all of which the matching can only partly absorb. Residual confounding by indication is real. A patient routed to surgery in 2018 is not the same patient routed to tirzepatide in 2024, and the difference is not fully captured by ICD codes and a propensity model.
The effect sizes are also unusually large. Odds ratios of 0.13 and 0.21 in real-world data should make any honest reader pause. They imply either a powerful biological effect, substantial residual confounding, or both. A randomized trial would settle it. Nobody is running one.
For practice, the analysis pushes the question of which intervention to offer first in obesity with IBD into open ground. For the field, it adds to a growing pile of evidence that GLP-1 drugs have anti-inflammatory effects which the weight-loss numbers do not fully describe. The cleanest version of that argument now comes from a comparison the field would not have predicted, against a comparator nobody expected to lose.