A new review of semaglutide trials reports a substantial average weight-loss advantage over placebo in adults without diabetes. Its more useful finding for readers is how much that advantage varied: the authors rated the certainty of the pooled weight-change estimate as low. The result supports a benefit in the studied populations, but it cannot supply one reliable percentage for every formulation, treatment period or person.
The review published on October 31 combined 10 randomized trials involving 7,131 participants. It searched evidence through February 28, 2026. This is a new synthesis of existing trials, not a newly completed trial or an approval announcement. Its focus is adults with obesity, or overweight plus a weight-related condition, without diabetes.
12.29 points is a comparison, not a personal target
For percentage change in body weight, the review reported a weighted mean difference of −12.29 percentage points between semaglutide and placebo, with a 95% confidence interval from −13.38 to −11.20. That analysis used 6,975 participants, a smaller denominator than the overall review. The negative sign indicates a greater reduction with semaglutide. It does not mean everyone taking the medicine lost exactly 12.29% of their starting weight.
A single trial shows why the distinction matters. In the original STEP 1 report, 1,961 adults without diabetes received weekly injected semaglutide 2.4 mg or placebo, with lifestyle support, for 68 weeks. Average weight changes were −14.9% and −2.4%, respectively; the reported estimated treatment difference was −12.4 percentage points. Total weight change in one group and the difference between groups answer separate questions.
That dose describes the research regimen, not a treatment instruction. The STEP 1 registry confirms that both groups received diet and physical-activity counselling. Its results therefore describe a supported trial setting. For the broader product and trial picture, see our semaglutide evidence guide.
Why a large collection of trials can still yield low certainty
The pooled weight-percentage result had an I² of 88%, a statistical signal of substantial variation between studies beyond sampling error. A narrow confidence interval around an average does not make the studies interchangeable. Nor does 88% describe the proportion of patients who responded or failed to respond.
The review’s supplementary GRADE table explains the low-certainty rating: the authors downgraded this outcome by two levels for very serious inconsistency. Removing one study at a time did not resolve the heterogeneity. This is a caution about the magnitude of the combined estimate, not a finding that all the randomized trials were worthless.
The abstract also reports a 95% prediction interval from −16.06 to −8.52 percentage points. Unlike the confidence interval around the pooled mean, this describes uncertainty about the true effect in another comparable study under the model’s assumptions. It is not a range within which an individual’s weight loss is expected to fall. The distinction helps explain why a precise-looking average can coexist with considerable variation.
The pooled result does not describe one treatment plan
The supplement lists different regimens and study durations. Its analysis table labels most weight-change results as treatment-policy estimates, while the earlier phase 2 study has a different label. STEP 1 explicitly counted effects regardless of treatment discontinuation or rescue interventions. We cannot assume that every trial handled these events identically simply because several share a statistical label.
These differences matter when a pooled number is used to compare pills with injections, or one dose with another. The review’s overall placebo comparison does not establish a head-to-head ranking of those options. Our guide to trial estimands explains how the treatment question changes when interruptions are handled differently; the semaglutide drug profile keeps the medicine’s identity separate from any one study result.
Stopping because of adverse effects deserves its own line
The review reported aggregate adverse-event discontinuation proportions of 6.0% with semaglutide and 3.0% with placebo, alongside a pooled risk ratio of 1.86. Those percentages summarize events across trials; they are not an individual forecast or a substitute for the weighted analysis. They show why tolerability belongs next to weight-loss results, rather than in a footnote.
For serious adverse events, the pooled risk ratio was 1.16, with a 95% confidence interval of 0.90 to 1.51. The review did not establish a clear difference, but that interval also leaves room for a clinically important increase in risk. Describing this as proof of equal safety would overstate the evidence. Our discontinuation and tolerability guide explains why stopping treatment, missing follow-up and serious adverse events are different outcomes.
What to take into an evidence discussion
The useful questions are specific: Which formulation and trial population are relevant? How long were participants followed? Did the estimate include people who stopped treatment? What were the tolerability findings? These questions are more informative than adopting a pooled percentage as a target.
We checked the complete review abstract, selected supplementary tables, the STEP 1 abstract and its registry record. The review’s full main paper was not available for inspection, so this article does not claim a complete audit of its methods or every included trial. The new synthesis adds context about variability and uncertainty; it does not decide which medicine, dose or monitoring plan is right for a reader.



