GLP-1 and muscle

What the trial record reports on discontinuation

This page reports figures published in randomised trials and one network meta-analysis, with the design context that makes them comparable or not. It does not say any compound is well tolerated or poorly tolerated, and it gives no advice about starting or stopping anything — that requires someone who can examine you.

Two different measurements share one word

In a trial report, discontinuation due to adverse events counts participants who stopped taking the study product because something happened to them. It is a tolerability measure, and it is meaningful only against the placebo column printed beside it.¹

A randomised withdrawal trial is something else entirely: everyone is treated first, then some are switched off treatment deliberately, and the outcome measured is what happens afterwards. STEP 4 and SURMOUNT-4 are both built this way, and neither is a study of why people stop.³⁵¹⁰

What the tolerability numbers actually say

STEP 1 randomised 1,961 adults with overweight or obesity without diabetes for 68 weeks. More participants receiving semaglutide than placebo discontinued treatment owing to gastrointestinal events: 59 (4.5%) versus 5 (0.8%). That figure covers gastrointestinal events specifically, in that population, over that window.¹⁸

SURMOUNT-1 randomised 2,539 adults over 72 weeks. Adverse events caused treatment discontinuation in 4.3%, 7.1% and 6.2% of participants across the three tirzepatide groups, against 2.6% with placebo. Three different numbers came out of one trial, which is itself a caution against quoting a single value for a compound.²

SELECT enrolled 17,604 people with pre-existing cardiovascular disease and overweight or obesity without diabetes, and followed them for a mean of 39.8 months. Adverse events leading to permanent discontinuation of trial product occurred in 1,461 participants (16.6%) receiving semaglutide and 718 (8.2%) receiving placebo. The higher figure is partly a longer trial in an older, sicker population — not a different drug behaving differently.⁴⁹

• Longer trials accumulate more discontinuations, so trial duration alone moves the number.
• Populations differ: a cardiovascular outcomes trial enrols people with more comorbidity than a weight-management trial.
• The placebo column moves too — 0.8% in STEP 1 and 8.2% in SELECT — which is why a treatment figure quoted on its own is uninterpretable.
• Some trials report all-cause discontinuation, some report adverse-event discontinuation, and some report a single symptom class.¹²⁴

The lead-in problem

STEP 4 reported that similar proportions discontinued treatment because of adverse events with continued semaglutide (2.4%) and with placebo (2.2%). Taken bare, that reads as near-perfect tolerability. It is not.³

STEP 4 randomised only participants who had already completed a 20-week run-in on the active drug. Anyone who could not tolerate it had left before randomisation, so the randomised population was pre-selected for tolerance. In the same trial, gastrointestinal events were still reported by 49.1% of those continuing semaglutide versus 26.1% on placebo.³

What the pooled evidence adds

A 2026 BMJ network meta-analysis of 262 randomised trials and 99,791 participants across 19 drugs assessed discontinuation because of adverse events among its 24 outcomes. It reported, with moderate to high certainty, that discontinuation for adverse events was highest with orforglipron, naltrexone–bupropion, liraglutide, phentermine–topiramate, cagrilintide–semaglutide and oral semaglutide, with risk ratios from 1.9 to 4.2 relative to lifestyle modification.⁷

Those are relative risks against a comparator, from trials of 12 to 172 weeks, and the authors' own conclusion is that larger weight-loss benefits were generally accompanied by greater harms and discontinuation. A ratio of that kind orders drugs; it does not tell an individual what their own experience will be.⁷

What the withdrawal trials measured instead

The STEP 1 extension followed 327 participants for a year after all treatment stopped at week 68. Mean weight loss to week 68 had been 17.3% with semaglutide and 2.0% with placebo; by week 120 those groups had regained 11.6 and 1.9 percentage points respectively, leaving net losses of 5.6% and 0.1%. All extension analyses were exploratory.⁶

SURMOUNT-4 randomised 670 participants who had completed a 36-week open-label lead-in. From week 36 to week 88 mean weight change was −5.5% with continued tirzepatide and +14.0% after the switch to placebo.⁵¹⁰

Both designs answer what follows a planned stop under trial conditions. Neither was designed to describe the people who stop in ordinary practice — for cost, for supply, for a side effect, or because they decided to — and no randomised trial of that population exists to quote.⁵⁶

Semaglutide is the compound with the widest published spread of discontinuation figures — 0.8% to 16.6% across the placebo and treatment columns of different trials — which is why a single quoted percentage for it should always be traced back to the trial and design it came from.

Common questions