Muscle loss on a GLP-1: what the body-composition studies actually measured

Before you read. This page summarises published body-composition research. It is educational, it is not advice about your treatment, and it cannot assess your muscle health. Crucially, none of the studies below used an oral semaglutide tablet — they used injectable drugs in obesity trial populations, and the differences in drug, dose, population and duration are exactly the differences that make one trial's numbers a poor prediction for another. Take any concern about strength, mobility or nutrition to your prescriber or a registered dietitian rather than adjusting anything on your own.

Quick answer

Three things are true at once, and skipping any of them produces a misleading answer.

The scans measure lean mass, and lean mass is not muscle. It is everything that is not fat — organs, bone, connective tissue, body water, including the water held in fat tissue itself. Losing lean mass on a DXA scan is not the same finding as losing skeletal muscle.

In the best-designed substudy, the split was the same with the drug and without it. In a body-composition substudy of the SURMOUNT-1 trial, roughly three-quarters of the weight lost was fat and about a quarter was lean mass — and that proportion held for the placebo group too. Whatever produced the ratio, it was weight loss rather than the medicine.

The published range is far wider than the quoted number. Across studies, reported lean-mass reductions have run from around 15% of total weight lost to between 40% and 60%, depending on the population, the drug and how the measurement was made.

The response with actual support behind it is unglamorous and is the same one that applies to any substantial weight loss: resistance training and enough protein, with function measured rather than assumed.

Keep strength notes on the same timeline as the dose →

What the semaglutide tablet label says about this

Almost nothing, and that is worth stating rather than filling in.

The Rybelsus and Ozempic tablets prescribing information contains no body-composition substudy, no DXA data, and no lean-mass figure. Muscle loss is not listed as an adverse reaction. The adverse reactions reported at 5% or more are gastrointestinal plus decreased appetite, and symptom by symptom against the label goes through what is and is not attributed to the drug.

That silence is not evidence of absence. It reflects what the trial programme measured, which was glycemic control in type 2 diabetes over 26 weeks and cardiovascular outcomes over four years. Body composition was not an endpoint, so the label has nothing to report.

Everything else on this page therefore comes from other drugs in other populations. That is a real limitation and it is carried through every section below rather than mentioned once and dropped.

Lean mass is not muscle

This is the definitional problem underneath the entire debate, and it is the reason two honest researchers can read the same scan differently.

DXA — dual-energy X-ray absorptiometry — separates the body into fat mass, lean soft tissue and bone. Lean soft tissue is a residual category. As one review puts it plainly, changes in lean mass may not always reflect changes in muscle mass, because lean mass includes not only muscle but also organs, bone, fluids, and the water held within fat tissue.

Two consequences follow.

Losing body water registers as losing lean mass. Someone eating less, carrying less glycogen and less fluid will show a lean-mass drop that has nothing to do with muscle fibre.

Shrinking fat tissue takes its water with it. The water inside adipose tissue is counted as lean, so a large reduction in fat mass mechanically reduces measured lean mass.

Which is why the more recent work reaches for magnetic resonance imaging instead, where muscle volume can be measured directly rather than inferred from a residual.

The substudy everyone quotes

The most-cited number comes from a body-composition substudy of SURMOUNT-1, and the design matters as much as the result.

What it was: 160 participants out of the 2,539 in the main trial — 124 on pooled tirzepatide doses, 36 on placebo — scanned by DXA at baseline and again at week 72. Participants were 73% female, with a mean weight of 102.5 kg and a mean body mass index of 38.0.

What it found: of the body weight lost, approximately 75% was fat mass and 25% was lean mass — and, in the sentence that gets left out of almost every summary, that split was the same for tirzepatide and for placebo. The proportions held across subgroups by sex, by age band, and across tertiles of how much weight had been lost.

We are not reproducing the magnitudes here, for the same reason this site does not print headline weight-change figures anywhere else. The proportion is the part that answers the question being asked.

Note also what this study is not. It is tirzepatide, a dual receptor agonist, in a population recruited for obesity, over 72 weeks, in a substudy of 160 people. It is not oral semaglutide, and lining its numbers up against a different drug's trial would be exactly the cross-trial reasoning that produces confident wrong answers.

The measurement that matters here is function, and function is a diary rather than a scan. Takely keeps a line a day beside your dose history, so "the stairs got harder in March" has a date attached to it. Open Takely in your browser →

Why the published range runs from 15% to 60%

If the answer were simply "a quarter," there would be no debate. There is a debate.

Reviewing the field, one summary describes the heterogeneity directly: in some studies lean-mass reductions ran between 40% and 60% of the total weight lost, while others reported around 15% or less. That is a fourfold spread in the headline statistic.

The causes are the ordinary ones, and each is worth knowing because each tells you which studies to trust for which question:

Population. Older participants, people with more advanced disease, and people who were already losing weight before enrolment do not behave the same.
The drug. Different molecules, different doses, different durations.
The measurement. DXA residual versus MRI muscle volume, and whether anything was measured about strength or function at all.
The comparison. Whether a control group was scanned, which is the difference between knowing the ratio and knowing whether the drug caused it.

When a single number is quoted at you without its population attached, that number came from somewhere in this spread and the choice of where was somebody's.

Adaptive, or a problem?

This is the live scientific question, and the current weight of expert opinion leans one way while explicitly naming who should not be reassured.

The argument for adaptive goes like this. Body mass and muscle mass move together in both directions — larger bodies carry more muscle because they carry more load. When contemporary evidence is read with MRI added, the reduction in muscle volume appears commensurate with what would be expected given age, disease status and the amount of weight lost. Alongside that, insulin sensitivity improves and fat infiltration within muscle decreases, which points to muscle quality improving even as volume falls. On that reading, the probability of losing strength or function is lower than the volume change alone would suggest.

The argument for caution is not a rebuttal of that; it is a statement about who the average conceals. The same authors name older age and prefrailty as factors that should influence who is a good candidate, because of sarcopenia risk. An adaptive change in a 45-year-old and the same percentage change in a frail 78-year-old are not the same event.

There is a methodological complaint running through all of this that is worth carrying away: the field has been measuring the wrong thing. What we mostly have is quantity. What matters clinically is function, mobility and strength, and the recommendation from these reviews is that future work measure those directly.

What the multi-society advisory recommends

In 2025 four organisations — the American College of Lifestyle Medicine, the American Society for Nutrition, the Obesity Medicine Association and The Obesity Society — published a joint advisory on nutrition alongside GLP-1 treatment. It lists muscle and bone loss among the challenges of these therapies and does not treat the adaptive argument as settling the practical question.

Its recommendations on this point are short and unsurprising:

Resistance training, as an explicit priority rather than general activity advice
Appropriate protein intake, in the context of a reduced overall calorie intake
Baseline and ongoing assessment including muscle strength, function and body composition
Attention to nutrient deficiencies generally, since intake falls

It also recommends registered dietitian counselling, which is the honest answer to most of the questions this page raises. Protein targets during pharmacological weight loss are individual, and a page cannot set yours.

What any of this means if you take a tablet

Here is where an honest page has to stop short of the answer people want.

There is no oral semaglutide body-composition substudy to cite. It is tempting to reason that a lower-exposure oral treatment for glycemic control produces smaller changes than a high-dose injectable used for obesity, and that reasoning is plausible — but plausible is not measured, and the whole point of the section above is that a number without its population attached is worthless.

What can be said within the label: Rybelsus is indicated for glycemic control and cardiovascular risk reduction in type 2 diabetes, and what this tablet is and is not licensed for covers that boundary. The trials were not weight-loss trials and the label sets no weight expectation.

What can be said generally: the mechanism in the studies above is weight loss, not the molecule — that is the finding from the placebo arm. Any treatment or circumstance producing comparable weight loss raises the same question, and the mitigation is the same regardless of which one it is.

What is worth tracking, and what is not

Three things are worth writing down. One is worth ignoring.

Strength, as a specific act. Not "I feel weaker" but the thing that changed — a staircase, a shopping bag, standing from a low chair without hands. Those are the closest thing to a functional measure available at home, and they are what the reviews say the field should have been measuring.

Protein, roughly. Not to a gram. Enough to notice a week where intake collapsed, which happens easily when appetite is suppressed.

Resistance training, as a yes or no per session. A count is more useful than an intention, particularly across a stretch of months.

What to ignore: a bathroom scale's body-fat percentage. Consumer bioimpedance devices estimate body composition from electrical resistance, which is heavily influenced by hydration — the same confound that makes lean mass a slippery measure in a research-grade DXA scan, in a much less controlled setting. Watching that number fluctuate will tell you about your fluid balance and read like a story about your muscle.

Keeping strength notes on the same timeline as the dose is what turns a vague impression into something a clinician can act on, and the seven notes worth keeping in a week covers the wider version of that habit.

The question that matters more than the percentage

One finding from the withdrawal literature reframes this whole debate, and it deserves to be the last word.

When treatment is stopped, weight comes back. In the STEP 1 extension, participants regained about two-thirds of the weight they had lost over the following year, and the cardiometabolic improvements largely reverted. What the withdrawal data showed covers that study and its limits in full.

Regained weight is not regained muscle. It arrives disproportionately as fat, which means the composition question is not only about the months on treatment — it is about the cycle. That is an argument for taking the resistance training seriously and, more than that, for treating "how long is this for" as a real question to ask your prescriber at the start rather than at the end.

Frequently asked questions

Does Rybelsus cause muscle loss?

The label does not list muscle loss as an adverse reaction and its trial programme did not measure body composition, so there is no label answer. The published body-composition studies used injectable drugs in obesity populations, and their findings cannot be transferred to a different drug at a different exposure in a different population.

How much of the weight lost on a GLP-1 is muscle?

The most-quoted figure is about a quarter, from a DXA substudy of SURMOUNT-1 — but that substudy measured lean mass rather than muscle, and found the same proportion in the placebo group. Across the wider literature, reported lean-mass reductions range from around 15% of total weight lost to between 40% and 60%.

Is lean mass the same as muscle?

No. Lean mass on a DXA scan is everything that is not fat: muscle, organs, bone, connective tissue and body water, including water held in fat tissue. This is why researchers have moved toward MRI-based muscle volume for questions specifically about muscle.

Is the muscle change harmful?

Current expert reviews describe it as adaptive in most people — commensurate with the age, disease status and weight loss involved, alongside improvements in insulin sensitivity and muscle fat infiltration that suggest better muscle quality. Those same reviews single out older age and prefrailty as reasons for caution, and note that strength and function have been measured far less than mass has.

How do I prevent muscle loss on a GLP-1?

The 2025 multi-society advisory names resistance training and appropriate protein intake alongside a reduced calorie intake, plus assessment of muscle strength, function and body composition. Individual protein targets belong to a registered dietitian rather than to a web page.

Should I use a body-fat scale to track this?

Consumer bioimpedance scales estimate composition from electrical resistance and are strongly affected by hydration, which is the same confound that complicates lean-mass measurement in research settings. Functional markers — stairs, carrying, standing from a chair — are more informative at home.

Primary sources and further reading

Studies were located through PubMed and each is linked by DOI above. This is an active research area and the picture is still changing; everything here was checked on 8 September 2026.