An estimated 1.6 million adults in Great Britain used a GLP-1 or dual GLP-1/GIP medicine for weight loss in the year to early 2025. The MHRA has strengthened UK product information for these medicines twice since January 2026. Despite that scale of use, the evidence on GLP-1 side effects is more uneven than the prescribing volume suggests. Not all of these effects sit on the same evidential footing. Some effects are firmly quantified by randomised trials. Others were flagged by pharmacovigilance databases, investigated at length, and largely set aside. A third group remains genuinely unresolved. Distinguishing between the three matters, because the clinical response to each is different.
The GLP-1 Side Effects Seen in Every Trial
Gastrointestinal symptoms dominate. A pooled analysis of the STEP 1 to 3 trials covered 2,117 participants on semaglutide 2.4 mg and 1,262 on placebo. Nausea affected 43.9 per cent of treated participants against 16.1 per cent on placebo. Diarrhoea affected 29.7 per cent against 15.9 per cent, vomiting 24.5 per cent against 6.3 per cent, and constipation 24.2 per cent against 11.1 per cent.
The severity distribution matters as much as the frequency. In that analysis, 99.5 per cent of gastrointestinal events were non-serious and 98.1 per cent were mild to moderate. Symptoms clustered during and shortly after dose escalation, then receded. Only 4.3 per cent of semaglutide-treated participants stopped permanently because of them.
A systematic analysis covered 38 placebo-controlled phase III and IV trials in people with type 2 diabetes. Nausea affected 19.3 per cent of participants on active treatment against 6.5 per cent on placebo. Vomiting affected 7.6 per cent against 2 per cent. Across randomised trials, 6.5 per cent of participants discontinued because of adverse events, compared with 3.6 per cent on placebo.
Relative risks vary by agent. A network meta-analysis of 39 randomised trials in people without diabetes put the relative risk of nausea at 2.95 for semaglutide, 2.90 for tirzepatide and 4.77 for the oral agent orforglipron. Tirzepatide carried the highest relative risk of vomiting, at 13.23. That result is notable because dual GIP and GLP-1 agonism was expected to improve tolerability. Preclinical work supports the idea, but comparative trial data has not borne it out.
Why These Effects Happen
Two mechanisms operate, and they are not mutually exclusive. GLP-1 receptor agonism slows gastric emptying and suppresses small intestinal motility. Separately, these drugs act on brain regions that sit outside the blood-brain barrier. These include the area postrema and the nucleus tractus solitarii, the structures implicated in medication-induced nausea generally. Signalling also reaches the brainstem indirectly through vagal afferent fibres.
This explains why the common effects track dose escalation rather than duration of treatment. It also explains why slower titration helps. A randomised open-label pilot published in 2025 tested a 16-week flexible titration schedule. It produced better adherence and fewer adverse events than the eight-week regimen specified on the label.
How These Effects Are Measured, and Why It Matters
Ryan Jalleh, Nicholas Talley, Michael Horowitz and Michael Nauck reviewed the class in the Journal of Clinical Investigation in February 2026. They identified a structural weakness in how these events are captured. Almost all trials rely on participant self-report without a standardised instrument. Terms are therefore interpreted differently between participants. Uncomfortable fullness may be recorded as nausea by one person but not another. Expectation effects compound the problem. Gastrointestinal symptoms are also common at baseline in the general population, and more common still in people with diabetes.
The authors argue for validated questionnaires of the kind regulators already mandate in irritable bowel syndrome trials. Until that happens, apparent differences between individual compounds should be treated with caution.
Risks That Have Been Investigated and Downgraded
Some of these risks entered public discussion on weak evidence and have since been reassessed. Acute pancreatitis is the clearest example. Case reports involving exenatide prompted searches of adverse event databases. Those searches suggested a substantially raised risk, and pancreatitis became an event of special interest across the class. Adjudicated trial data told a different story. The LEADER trial recorded 18 cases among 4,668 participants on liraglutide against 23 among 4,672 on placebo. Meta-analyses of cardiovascular outcomes trials have since, in the reviewers’ phrasing, “dispelled” the causal claim.
Part of the explanation is circularity in the diagnostic criteria. Acute pancreatitis is diagnosed on two of three features: severe upper abdominal pain radiating to the back, raised amylase or lipase, and characteristic imaging. GLP-1 therapy independently produces abdominal symptoms in a large proportion of patients and elevates pancreatic enzymes in the majority. Two of the three criteria can therefore be satisfied by the treatment itself.
That is not the end of the matter. In January 2026 the MHRA strengthened product information across the class. The agency had reviewed 1,296 Yellow Card reports of pancreatitis received between 2007 and October 2025. Of those, 19 were fatal and 24 involved necrotising pancreatitis. Set against roughly 25.4 million packs dispensed in the UK over five years, the reported rate is very low, but the severity of those outcomes prompted the Commission on Human Medicines to act. Alison Cave, the MHRA’s Chief Safety Officer, described the risk of developing these severe effects as “very small” while urging awareness of the symptoms.
Psychiatric effects have followed a similar path. A systematic review covered 80 randomised trials involving 107,860 participants. It found no association with major depression, suicide or psychosis, and reported improved mental health related quality of life.
The GLP-1 Side Effects That Remain Unresolved
Four areas are genuinely open, and they account for most of the current uncertainty about serious adverse effects.
Thyroid cancer sits first. Rodent studies showed C-cell proliferation. Most human medullary thyroid carcinomas express the GLP-1 receptor, although healthy human C cells largely do not. French health insurance data indicated a raised risk of medullary thyroid carcinoma, at a hazard ratio of 1.78. The figure for thyroid carcinoma generally was 1.58. A Scandinavian cohort study using national cancer registries found no association, at a hazard ratio of 0.93. Personal or family history of medullary thyroid carcinoma or multiple endocrine neoplasia type 2 remains a contraindication.
Optic neuropathy has produced the most contested signal of the past two years. Cohort studies of non-arteritic anterior ischaemic optic neuropathy have reported hazard ratios ranging from 1.39 to above 4. One Danish analysis of 424,152 people with type 2 diabetes found a doubling of five-year risk. Several other studies found no association at all. Absolute incidence remains low. Regulators have acted regardless. The MHRA updated semaglutide product information in February 2026. It described an approximate doubling of relative risk, corresponding to roughly one additional case per 10,000 patients treated each year. The European Medicines Agency’s Pharmacovigilance Risk Assessment Committee classified the condition as a very rare adverse effect.
Gallbladder disease is better quantified. A systematic review of 55 placebo-controlled trials found a raised risk of cholelithiasis, at a relative risk of 1.46. There was no increase in cholecystitis, cholangitis or pancreatitis.
Retained gastric contents raises a practical procedural question rather than a chronic one. One study found retained contents in 56 per cent of GLP-1 users before anaesthesia, against 19 per cent of non-users, although the treated group contained more people with type 2 diabetes and obesity. Three systematic reviews reported substantially higher odds of endoscopy being terminated early. None identified an increase in aspiration pneumonia. Multi-society guidance recommends an individualised approach to withholding treatment rather than a fixed pre-procedure rule.
Lean mass is the fourth. Fat-free mass accounted for 39 per cent of total weight loss with semaglutide and 25 per cent with tirzepatide. That measure is a crude proxy for muscle, and no study has yet assessed functional consequences.
Side Effects Outside the Trial Setting
Trial discontinuation figures understate real-world behaviour by a wide margin. A Danish registry study presented at the 2025 European Association for the Study of Diabetes meeting followed 77,310 adults without diabetes who started semaglutide for weight loss. Within three months, 18 per cent had stopped. Within a year, 52 per cent had. Reviews of observational data put first-year discontinuation between 20 and 50 per cent. SELECT was the cardiovascular outcomes trial in people with obesity and established cardiovascular disease. There, 16.6 per cent of participants permanently discontinued semaglutide against 8.2 per cent on placebo. Gastrointestinal disorders accounted for 10.0 per cent against 2.0 per cent.
Cost and access explain part of that gap. Tolerability explains a substantial share of the rest. That is why the measurement problem is not merely academic. GLP-1 side effects are a leading reason people stop taking a medicine intended for indefinite use. Quantifying them precisely is therefore a question of effectiveness, not comfort.
The Gaps Regulators Acknowledge
Three Cochrane reviews commissioned by the World Health Organization were published in late 2025. All three found clinically meaningful weight loss across liraglutide, semaglutide and tirzepatide. The reviewers judged the evidence on longer-term outcomes and side effects limited or uncertain, with much of it industry funded. The WHO guideline that followed in December 2025 recommends this class for obesity conditionally, on moderate to low certainty evidence, and specifies periodic monitoring of adverse events as part of chronic care.
For a class now taken by millions of people, with a substantial pipeline of next-generation agents behind it, that is a candid assessment of how much remains unknown.
The Practical Summary
Three categories, three different clinical responses. Gastrointestinal effects are common, predictable, dose-related and usually transient, and they drive most discontinuation. Pancreatitis and psychiatric harm have been examined thoroughly, and the causal claims have not held. Regulators have nonetheless strengthened warnings on rare severe pancreatitis outcomes. Thyroid, ocular, biliary and body composition questions remain open. Any confident claim about GLP-1 side effects, in either direction, is worth checking against which of those three categories it belongs to.
