GLP-1 and muscle

Gastrointestinal effects: what the trials actually measured

This page reports measured frequencies from randomised trials, approved labelling and one comparative cohort study, with the comparison groups attached. It makes no claim that any compound is safe or unsafe, and it does not advise on managing any symptom — that requires someone who can examine you.

The label tables, with their denominators

The Wegovy label tabulates adverse reactions in adults from pooled trials covering 2,116 participants on the drug and 1,261 on placebo. Nausea was reported by 44% versus 16%, diarrhoea 30% versus 16%, vomiting 24% versus 6%, constipation 24% versus 11% and abdominal pain 20% versus 10%.¹

The Zepbound table is stratified into three treatment columns against 958 placebo participants. Across those columns nausea ran 25% to 29% versus 8% on placebo, diarrhoea 19% to 23% versus 8%, vomiting 8% to 13% versus 2% and constipation 11% to 17% versus 5%.²

What the trials reported in their own words

STEP 1, which randomised 1,961 adults for 68 weeks, reported nausea and diarrhoea as the most common adverse events with semaglutide, described them as typically transient and mild-to-moderate in severity, and reported that more participants on semaglutide than placebo discontinued treatment owing to gastrointestinal events — 59 (4.5%) versus 5 (0.8%).³

SURMOUNT-1, which randomised 2,539 adults for 72 weeks, reported that the most common adverse events with tirzepatide were gastrointestinal, mostly mild to moderate, and occurred primarily during the escalation period at the start of treatment.⁴

STEP 4 gives the clearest single frequency for the class as a whole: gastrointestinal events were reported in 49.1% of participants continuing semaglutide versus 26.1% on placebo. That is in a population that had already completed a 20-week run-in on the drug, so it describes people who had, by definition, tolerated it that far.⁵

One head-to-head comparison

SURPASS-2 randomised 1,879 people with type 2 diabetes to tirzepatide or semaglutide over 40 weeks — no placebo arm, so the comparison is between the two drugs rather than against nothing. Nausea was reported by 17% to 22% across the tirzepatide groups and 18% with semaglutide, diarrhoea 13% to 16% and 12%, vomiting 6% to 10% and 8%.⁶⁹

Those ranges overlap. A trial designed and powered for glycaemic control is not powered to resolve small differences in symptom frequency, and the authors did not present these as a comparative tolerability conclusion.⁶

The uncommon conditions, and why they are harder

Pancreatitis, gastroparesis and bowel obstruction are too infrequent for a weight-management trial to count reliably, so the published estimates come from claims data. Sodhi and colleagues followed new users of semaglutide (613) and liraglutide (4,144) prescribed for weight loss against bupropion–naltrexone users (654) in a US claims database.⁷

Compared with bupropion–naltrexone, GLP-1 agonist use was associated with pancreatitis (adjusted hazard ratio 9.09, 95% CI 1.25–66.00), bowel obstruction (4.22, 95% CI 1.02–17.40) and gastroparesis (3.67, 95% CI 1.15–11.90), and not with biliary disease (1.50, 95% CI 0.89–2.53). Every one of those intervals is wide, and the pancreatitis estimate rests on 2 events among semaglutide users, 71 among liraglutide users and 1 in the comparator group.⁷

The study is also observational. People prescribed a GLP-1 agonist differ from people prescribed bupropion–naltrexone in ways the authors adjusted for and in ways nobody can adjust for; the authors computed E-values to test how strong an unmeasured confounder would have to be, which is a strength of the paper and not a substitute for randomisation.⁷

What the pooled evidence says about the class

The 2026 BMJ network meta-analysis of 262 trials and 99,791 participants found gastrointestinal events most increased with naltrexone–bupropion, oral semaglutide, orforglipron and tirzepatide, with risk ratios from 3.1 to 4.2 relative to lifestyle modification, and concluded generally that larger weight-loss benefits were accompanied by greater harms and discontinuation.⁸

A risk ratio orders drugs against a common comparator. It is not a probability for a person, and the trials pooled ran from 12 to 172 weeks in populations selected to enter them.⁸

The same symptom names, in a much weaker source

Querying FDA's adverse event database for reports naming semaglutide returns the same words at the top of the list — nausea 943, vomiting 835, diarrhoea 589 in the extract dated 28 April 2026 — but these are counts of spontaneous, unverified reports with no denominator, and no rate can be computed from them.¹⁰

The agreement between the two sources is worth noticing and worth not over-reading. It tells you the reports are about the same phenomena the trials measured. It does not add any precision to the measurement, and where the two are quoted together the trial figure is the one carrying the information.¹⁰¹

Semaglutide has the fullest published gastrointestinal record of any compound on this site: label tables with denominators, several randomised trials reporting event frequencies, and a comparative cohort for the uncommon conditions.

Common questions