Semaglutide’s highest-dose weight-loss trial put a small group of its participants inside an MRI scanner, and the results speak to the question that trails every GLP-1 drug: what happens to muscle. Among 55 people drawn from the 1,407-person STEP UP trial, fat volume fell sharply over 72 weeks, thigh muscle volume fell by an amount too small to separate from chance, and the fat marbled inside the thigh muscle fell measurably. The number of times participants could stand up from a chair in 30 seconds did not drop.
What the scanner actually measured
STEP UP randomised adults with a body mass index of 30 or above to once-weekly semaglutide 7.2 mg, semaglutide 2.4 mg or placebo for 72 weeks. This prespecified secondary analysis covers only the participants with a valid baseline MRI scan: 49 on semaglutide (43 at 7.2 mg, six at 2.4 mg) and six on placebo.
Comparing the pooled semaglutide group against placebo, the estimated treatment differences in absolute volume change were −11.1 litres of adipose body tissue (95% confidence interval −16.5 to −5.7; p < 0.0001) and −1.5 litres of visceral adipose tissue, the fat packed around the organs (−2.9 to −0.1; p = 0.04). Both intervals sit entirely below zero, so the direction is not in doubt.
The muscle numbers behave differently. Lean body tissue fell by 1.7 litres more on semaglutide than on placebo, but the interval runs from −4.0 to +0.7 (p = 0.16) — it crosses zero, so this analysis cannot say there was a real difference. Thigh skeletal muscle volume fell by 1.1 litres, interval −2.3 to +0.1 (p = 0.07), which also crosses zero. The one muscle measure that separated cleanly went the favourable way: thigh muscle fat infiltration fell by 0.92 percentage points more on semaglutide (−1.41 to −0.42; p = 0.001). That is a measure of muscle quality rather than muscle quantity — less fat deposited within the muscle tissue itself. On the 30-second sit-to-stand test, the groups averaged 13 repetitions at baseline and 15 at week 72.
Why this lands in the middle of an argument
The claim that GLP-1 drugs strip muscle along with fat has been the most persistent criticism of this class, and the evidence behind it is mostly indirect: bioelectrical impedance readings, DXA scans, and the observation that any large weight loss takes some lean tissue with it. MRI is a more exacting instrument than impedance, and it can separate muscle volume from the fat inside the muscle — a distinction impedance cannot make. This analysis found the volume signal weak and the quality signal clear, in the opposite direction to the worry. It does not settle the argument, but it is the kind of measurement the argument has been missing. Our overview of what the trials show on semaglutide covers the main STEP UP results, which were published separately.
Note also what physical function did here. Sit-to-stand repetitions rose from 13 to 15 across the groups over 72 weeks. Whatever happened to muscle volume in these participants, it did not show up as fewer repetitions on the one functional test reported.
What is not yet known
This is a very small substudy inside a very large trial: 55 of 1,407 participants had a usable baseline scan, and only six of them were on placebo. A six-person comparison group is why three of the five confidence intervals above are as wide as they are, and why two of them cross zero. A real difference in muscle volume could easily hide inside an interval running from −4.0 to +0.7 litres; so could no difference at all. The sit-to-stand figures are reported as a pooled average across semaglutide and placebo, not as a between-group difference, so they do not establish that function was better on the drug than off it. Several authors are employees of Novo Nordisk, which makes semaglutide and funded the trial, and the imaging analysis was run by AMRA Medical. And 72 weeks is a short window for a question — what sustained weight loss does to muscle over years — that is really about decades.
Sources: Hjelmesæth J, Bhat S, Garvey WT, et al. “Effect of Semaglutide on Thigh Skeletal Muscle Volume, Fat Infiltration and Physical Function in People With Obesity: Secondary Analysis of STEP UP.” Diabetes, Obesity and Metabolism, published online 20 September 2026. doi:10.1111/dom.71323. Trial registration NCT05646706.
