Semaglutide meta-analysis: what it adds

A systematic review and meta-analysis published in Medicine on 31 July 2026 pooled four randomised trials of subcutaneous semaglutide in adults with overweight or obesity and without type 2 diabetes. Across 3,613 participants, the pooled mean difference in percentage body-weight change was -11.85% against placebo (95% CI -12.81 to -10.90)^[1]. That is almost exactly the number a single trial, STEP 1, reported in 2021^[2]. So the useful question here is not whether semaglutide reduces weight in this group. It is what a fresh pool of four already-published trials does and does not add.

What the pooled analysis reports

The authors followed PRISMA 2020 and searched PubMed/MEDLINE, Embase, Scopus and CINAHL from inception to March 2025 for randomised controlled trials of subcutaneous semaglutide in adults with overweight or obesity without type 2 diabetes. Four trials met the eligibility criteria. The primary outcome was percentage change in body weight; the secondary outcomes were gastrointestinal adverse events and treatment discontinuation^[1].

On efficacy, the pooled mean difference was -11.85% versus placebo. On tolerability, gastrointestinal adverse events (nausea, vomiting, diarrhoea and constipation) were significantly more frequent on semaglutide, and discontinuation because of adverse events was more than twice as likely (risk ratio 2.62, 95% CI 1.70 to 4.03). Serious events such as acute pancreatitis and cholelithiasis remained rare^[1]. None of that contradicts the safety profile already on the semaglutide page.

Why the headline number looks familiar

STEP 1 randomised 1,961 adults with overweight or obesity and no diabetes to semaglutide 2.4 mg weekly or placebo for 68 weeks. Mean body-weight change was -14.9% on semaglutide versus -2.4% on placebo, an estimated treatment difference of -12.4 percentage points (95% CI -13.4 to -11.5)^[2]. The new pooled estimate of -11.85% (95% CI -12.81 to -10.90) sits inside that interval. One trial from five years ago and a four-trial pool from 2026 are saying the same thing.

That is not a coincidence, and it is not a flaw in the arithmetic. Standard inverse-variance meta-analysis weights each trial by precision, so that, in the Cochrane Handbook's words, "larger studies, which have smaller standard errors, are given more weight than smaller studies, which have larger standard errors"^[3]. A pool of 3,613 participants is under twice the enrolment of STEP 1 alone, so the pooled answer tracks its biggest contributor closely. Readers arriving at the semaglutide overview from a "new meta-analysis confirms" headline should read that as consistency, not fresh evidence.

Two things worth flagging to anyone citing this

First, the abstract does not name the four included trials, state their doses, or give their durations. A reader trying to work out how much of this pool is the STEP programme, and therefore how much they have already read, cannot do it from the abstract. The full text sits behind the journal's paywall.

Second, the literature search ran to March 2025 but the paper appeared in July 2026. Roughly sixteen months of trial output in the fastest-moving area of GLP-1 receptor agonist research falls outside the search window. Nothing in that gap is likely to overturn a -11.85% estimate, but the paper is a snapshot of an older evidence base than its publication date suggests. That matters most when setting it beside our tirzepatide coverage, where the trial programme has moved since early 2025.

The number people misread

A pooled mean difference of -11.85% is not the weight an individual loses. It is the gap between the semaglutide arm and the placebo arm. Placebo arms in obesity trials also lose weight, because both arms usually receive the same structured lifestyle support: in STEP 1 the placebo group lost 2.4% of body weight on its own^[2]. The arm-level figure and the placebo-subtracted figure are different quantities, and marketing copy tends to blur them.

The other thing a mean hides is spread. In STEP 1, 86.4% of the semaglutide group lost at least 5% of body weight, 69.1% lost at least 10%, and 50.5% lost at least 15%^[2]. Half the group cleared 15% and half did not. A pooled average compresses that spread into one figure: the right summary for a guideline committee, the wrong one for predicting any single person's result.

Where pooling actually earns its keep

The discontinuation result is the more useful half of this paper. Single trials are designed to detect a weight-loss difference, not to give precise estimates of how many people quit because of side effects. STEP 1 reported 4.5% of the semaglutide group discontinuing for gastrointestinal reasons against 0.8% on placebo^[2]. Pooling four trials tightens that kind of estimate, and a risk ratio of 2.62 with an interval of 1.70 to 4.03 is a reasonably firm statement about tolerability-driven dropout.

It is still a ratio, and the base rate matters: doubling a small absolute risk is not the same proposition as doubling a large one. The abstract gives the ratio, not the pooled absolute rates. The European Medicines Agency already lists nausea, vomiting, diarrhoea, constipation, headache and abdominal pain among the side effects that may affect more than 1 in 10 people taking Wegovy, authorised for adults with a BMI of 30 or above, or 27 and above with a weight-related condition^[4]. The pooled signal is consistent with that label rather than a new warning, and our semaglutide regulatory summary reflects the same position.

What four trials cannot settle

The authors say plainly that long-term safety data require further investigation. Four trials pooled to March 2025 do not speak to durability beyond the trial windows, to what happens after people stop, or to how semaglutide performs head to head against tirzepatide or retatrutide. The Cochrane Handbook also cautions that "there is much uncertainty in measures such as I² and Tau² when there are few studies", and that "investigations of heterogeneity when there are very few studies are of questionable value"^[3]. Four is few.

Our reading: this is a confirmation, not a development. Cite it for the stability of the effect size in non-diabetic obesity and for the discontinuation ratio. Do not reach for it to answer what semaglutide does over five years, how it compares with newer molecules, or what happens when treatment ends. For those, start with the semaglutide page.

This article is educational and is not medical advice. Semaglutide is a prescription-only medicine everywhere we cover. If you are considering it or already using it, the conversation about whether it suits your history belongs with a qualified healthcare professional.