Two symptoms sit next to each other on every one of these labels, and they behave nothing alike. In a network meta-analysis of 39 reports covering 33,354 adults with overweight or obesity and without diabetes, semaglutide raised the risk of belching by a factor of 7.88 and the risk of abdominal distension by a factor of 1.42.[1] Same drug, same trials, same analysis, a five-fold gap between the two. The one that people describe as the defining discomfort of the first months is the one the data barely separates from placebo.
What the incidence tables actually show
The semaglutide 2.4 mg weight-reduction pool reported abdominal distension in 7% of 2,116 treated adults against 5% of 1,261 on placebo — a two-point gap, the narrowest of any gastrointestinal entry in that table, where nausea runs 44% against 16%.[2] The tirzepatide weight-reduction pool recorded 3%, 3% and 4% across 5, 10 and 15 mg against 2% on placebo.[3] Liraglutide 3 mg came in at 4.5% of 3,384 treated adults against 3.0% of 1,941 on placebo.[4] Oral semaglutide reported 3% at 14 mg, 2% at 7 mg and 1% on placebo.[5]
The tirzepatide diabetes label is the one that gives the game away. Across placebo and the three maintenance doses it reports abdominal distension at 0.4%, 0.4%, 2.9% and 0.8%.[6] The lowest dose matches placebo exactly, the middle dose is seven times either, and the highest dose falls back to roughly a fifth of the middle one. A symptom produced by drug exposure does not trace that shape. Noise does.
The pooled estimate, and how few agents reach it
Eleven trials in the network meta-analysis contributed distension data. Semaglutide was the only agent to reach statistical significance, at a relative risk of 1.42 (95% CI, 1.13 to 1.79, p = 0.0026). Liraglutide came out at 1.57 (0.84 to 2.90) and exenatide at 6.56 (0.83 to 52.15), neither significant.[1] An interval running from 0.83 to 52 is not a finding about exenatide; it is a statement that almost nobody in those trials reported the symptom either way.
Set 1.42 against the other outcomes the same paper measured in the same population — vomiting at 4.21 for semaglutide, nausea at 2.95, upper abdominal pain at 2.14, belching at 7.88 — and distension is the floor of the whole panel.[1] That ranking is the practical content of this page. Someone whose main complaint after six weeks is a tight, distended abdomen has the gastrointestinal symptom least likely to be caused by the injection, and the most likely to have been present before it.
Bloating and distension are two different measurements
Part of the gap is definitional. The 2025 European consensus on this symptom pair, developed through a Delphi process across national societies, defines abdominal distension as an objective, visible increase in abdominal girth and bloating as a feeling of fullness and discomfort, and states explicitly that bloating may occur with or without distension.[7] Every label figure above counts distension — the reported sign. None of them counts the sensation, which is what the word bloating means when a patient uses it.
That matters in both directions. It means the small percentages above almost certainly undercount the discomfort people experience. It also means the complaint routinely gets attributed to a mechanism that has never been shown to produce it, because the sensation of fullness after a small meal is what an appetite-suppressing drug is supposed to create. The related question of stool frequency and the interventions for it sit in the constipation article, and belching has its own separate and much stronger signal in the eructation write-up.
The same consensus describes what actually produces the symptom when no drug is involved, and it is not one mechanism. Visceral hypersensitivity, abdomino-phrenic dyssynergia — the diaphragm descending and the abdominal wall relaxing in response to gas that a normal gut would accommodate without any visible change — intestinal dysmotility and dysbiosis are all named as contributors, and functional bloating frequently overlaps with other disorders of gut-brain interaction.[7] The interventions the consensus lists follow from that: dietary modification including lactose-limiting and low-FODMAP approaches, probiotics, antispasmodics such as peppermint oil, rifaximin, secretagogues, neuromodulators, and biofeedback, with cognitive behavioral therapy and hypnotherapy where the symptom sits alongside irritable bowel syndrome. None of those is a GLP-1-specific remedy, which is the point: a symptom with a relative risk of 1.42 is largely being treated as the condition it would have been without the injection.
The mechanism everyone names has been measured
Delayed gastric emptying is the explanation offered for this symptom everywhere, and it exists. A 2024 systematic review and meta-analysis pooled the studies that quantified it. Across five scintigraphy studies covering 247 participants, the mean gastric half-emptying time was 138.4 minutes (95% CI, 74.5 to 202.3) on a GLP-1 receptor agonist against 95.0 minutes (54.9 to 135.0) on placebo — a pooled difference of 36.0 minutes (17.0 to 55.0, p < 0.01).[8]
The same review found something the first figure conceals. Ten studies covering 411 participants used the acetaminophen absorption test, which reflects liquid emptying, and found no significant delay by either peak concentration or area under the curve. Agent, duration of treatment and short- versus long-acting mechanism did not change the result. The authors concluded that a delay of roughly 36 minutes is of limited magnitude relative to standard fasting periods.[8] The periprocedural version of that argument is in the anesthesia article.
The delay also does not hold. A pharmacology study comparing tirzepatide, dulaglutide and placebo found that a single dose delayed gastric emptying in participants with and without type 2 diabetes, and that the effect diminished after multiple doses in healthy participants — with only a residual delay remaining in patients with diabetes on a full escalation schedule.[9] So the mechanism attenuates with repeated exposure, in the same weeks that the trial data say gastrointestinal symptoms are at their peak. Whatever produces month-three distension, a 36-minute emptying delay that is already fading is a poor candidate for it. The escalation window those weeks belong to is mapped in the titration planner.
Where the line actually sits
The European consensus sets a workable boundary: in the absence of alarm features, and with an unremarkable history and examination, laboratory, imaging and endoscopic testing are unnecessary.[7] Distension that comes and goes with meals, eases overnight and coexists with the appetite suppression the drug is meant to produce is the ordinary version.
The version that is not ordinary appears in the postmarketing sections of the tirzepatide labels, which list ileus, intestinal obstruction and severe constipation including fecal impaction.[3] Those are reported voluntarily from a population of uncertain size, so they carry no incidence at all — but they define the shape of the emergency. A distended abdomen that will not pass stool or gas, that comes with vomiting or persistent severe pain, or that is firm and tender rather than merely tight, is an obstruction picture and needs same-day assessment rather than a smaller meal. Where distension persists alongside vomiting of food eaten many hours earlier, the question becomes the one covered in the gastroparesis article.
What the compounded versions do not tell you
No figure above was produced by a compounded product. The trials that generated them used FDA-approved drug at labeled doses, and most of the sellers reviewed on this site 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. The nearest available reading is a pharmacovigilance analysis of 81,078 GLP-1 reports, 707 of them compounded, which found higher reporting odds for abdominal pain at 2.84 (95% CI, 2.29 to 3.49) and for hospitalization at 2.35 (1.94 to 2.83).[10] Abdominal distension is not among the outcomes it reports, and a spontaneous database has no denominator, so neither the presence nor the absence of a signal there is an incidence.
What this leaves
Distension is real, it is uncomfortable, and it is the gastrointestinal complaint on this class with the weakest evidence of being drug-caused: two percentage points over placebo on the biggest label, a pooled relative risk of 1.42 for the one molecule that reached significance, and a dose curve on a second label that runs backwards. The mechanism usually cited for it is a 36-minute delay that fades with repeat dosing and does not show up in liquid emptying at all. The wider tolerability picture is in what the trials recorded.
None of that makes the symptom imaginary. It makes the attribution uncertain, which changes what to do about it: a complaint that predates the injection, or that behaves like a disorder of gut-brain interaction, has a management literature of its own and does not necessarily require holding a dose to address.