GLP-1 drugs slow the gut. That is not a side effect. It is part of how they work. By holding food in the stomach longer, they blunt appetite and flatten the blood-sugar spike after a meal. The same slowdown is why the drugs picked up a frightening warning. A new study of nearly 300,000 people with type 2 diabetes went looking for the feared consequence, a blocked or paralyzed bowel, and did not find it.
In September 2023 the FDA added ileus, a stall of the intestines that stops food and fluid from moving through, to the label for semaglutide ↗, the drug sold as Ozempic and Wegovy. The change ↗ came from voluntary post-marketing reports, and the agency was explicit that it could not say how often ileus happens or whether the drug causes it. That is the gap this study set out to fill.
The work, published online August 9 in Diabetes, Obesity and Metabolism ↗, used a Japanese commercial claims database called DeSC. The authors took adults with type 2 diabetes who newly started either a GLP-1 receptor agonist (the semaglutide class) or an SGLT2 inhibitor (a different diabetes drug that works on the kidney) between April 2016 and February 2024, and compared the two groups. Using one drug class as the yardstick for the other, rather than untreated people, is the cleaner design here. Both groups are patients sick enough to need a second diabetes drug, which strips out much of the difference between people who take medicine and people who do not.
The primary question was narrow and serious. A bowel obstruction or ileus bad enough to need a tube to decompress the gut or an operation, within one year of starting the drug. Among 298,124 eligible people, 25,779 started a GLP-1 drug and 272,345 started an SGLT2 inhibitor. After statistical weighting to balance the groups, the one-year rate was 0.14 percent on GLP-1 drugs versus 0.17 percent on the comparator. In plain terms, about 1 in 700 versus about 1 in 590, a difference of three hundredths of a percentage point that could easily be noise. The hazard ratio was 0.80, meaning the GLP-1 group's risk landed about 20 percent lower, but the confidence interval ran from 48 percent lower to 24 percent higher (0.52 to 1.24). That range crosses straight through no difference.
The secondary outcomes told the same story. Small bowel obstruction, small bowel obstruction needing surgery, paralytic ileus, and colonic volvulus (a twist of the colon) each showed no meaningful difference between the two drug classes.
What the null does and does not settle
A clean null is worth publishing, but it is worth reading carefully. This one rules out a large increase in bowel obstruction from GLP-1 drugs. It does not rule out a small one. The event is rare, roughly one in six hundred over a year, so even a database this size has limited power, and the wide confidence interval is the honest signature of that. A modest bump in risk could hide inside it.
The other caveats are the usual ones for claims data. The diagnoses come from billing codes, not chart review, so milder cases that never reached a tube or an operating room are invisible. The follow-up is one year, which catches an early problem and misses a slow one. The population is Japanese patients with type 2 diabetes, leaner on average than the American cohorts where the ileus reports piled up. And SGLT2 inhibitors are not inert. They are the comparator precisely because they are a fair match, not because they do nothing.
It joins a run of observational nulls on GLP-1 safety, from no melanoma signal in 390,000 matched pairs ↗ to no pregnancy-hypertension signal from early use ↗. Each is reassuring within its window, and each is unable to close the door on a small effect.
What this study does deliver is scale against anecdote. The ileus warning rests on a few dozen spontaneously reported cases worldwide. This is a quarter-million people run through the same question with a control group, and the alarming version of the story does not survive the comparison. Semaglutide and the newer dual-hormone drug tirzepatide ↗ both slow the gut. Neither, on this evidence, appears to be jamming it.