Do GLP-1 drugs cause disproportionate muscle loss?

People do lose lean mass while taking GLP-1-based weight-loss drugs. That part is visible in multiple human body-composition studies. The unresolved question is whether the loss is unusually large, whether it is actually skeletal muscle and whether it reduces strength or function.

The best current answer is narrower than either side of the argument. Fat mass usually falls more than lean mass. In one randomized tirzepatide substudy, the share of weight loss recorded as lean mass was about the same as with placebo. But the largest studies mostly used DXA, which does not isolate contractile muscle, and they rarely measured strength or physical performance.

So the evidence does not establish class-wide, disproportionate muscle wasting. It also does not prove that muscle is preserved.

One phrase hides four different outcomes

Lean mass, muscle mass, muscle quality and muscle function are not four names for the same thing.

DXA divides the body into fat, bone mineral and lean soft tissue. Its lean compartment includes skeletal muscle, but also organs, body water and other non-fat tissue. Hydration and stored glycogen can move the number. A decline in DXA lean mass is therefore a real body-composition change, but not a kilogram- for-kilogram measure of lost contractile muscle.

MRI and CT can isolate muscle volume or cross-sectional area and can estimate fat within muscle. Those scans get closer to muscle quantity and quality, but size still cannot tell whether a person became weaker. Handgrip, knee-extension force, gait and other performance tests answer another question again.

This measurement ladder matters because sarcopenia is not diagnosed from a single fall in whole-body DXA lean mass. None of the pivotal substudies below was a class-wide trial of sarcopenia.

Semaglutide made lean mass a larger share of a smaller body

The STEP 1 trial randomized 1,961 adults without diabetes to weekly semaglutide 2.4 mg or placebo, both with lifestyle intervention. An exploratory DXA analysis covered just 140 participants at nine sites: 95 received semaglutide and 45 placebo.

At 68 weeks, the semaglutide group in that substudy had lost 15.0% of body weight. Total fat mass fell 19.3%, visceral fat 27.4% and total lean body mass 9.7%. Because fat fell faster, lean mass rose by 3.0 percentage points as a share of body weight.

Both statements are true at once: the participants had less absolute lean mass and a higher lean-mass proportion. Reporting only the proportion turns a loss into apparent preservation. Reporting the absolute loss as muscle wasting turns a broad DXA compartment into a functional diagnosis. The substudy did not measure direct muscle mass, strength, gait or sarcopenia.

It was also an exploratory slice of the full trial, dominated by women and not designed as a definitive muscle study. Novo Nordisk funded STEP 1 and employed several authors.

Tirzepatide’s randomized comparison argues against an unusual share

The SURMOUNT-1 DXA substudy provides a cleaner proportional comparison. The main trial randomized 2,539 adults without diabetes; 160 had usable baseline and week-72 DXA scans. Of those, 124 received one of three tirzepatide doses and 36 received placebo.

Pooled across tirzepatide doses, mean body weight fell 21.3%, fat mass 33.9% and lean mass 10.9%. On placebo, the corresponding changes were 5.3%, 8.2% and 2.6%.

About three quarters of the lost weight was fat and one quarter lean mass in both groups. In this small substudy, tirzepatide produced far more total weight loss, including more absolute lean loss, but not a larger lean share than placebo.

That result is evidence against the claim that tirzepatide necessarily strips an unusually high proportion of lean tissue. It is not proof of zero risk. The substudy was small, 73% of participants were women, and it did not test muscle strength or performance. Eli Lilly funded it; six listed authors were Lilly employees and shareholders, while external authors disclosed industry ties.

Retatrutide shows the same boundary over 36 weeks

A Phase 2 retatrutide substudy enrolled 189 adults with type 2 diabetes from a 281-person randomized trial. Baseline DXA data were available for 155 people; only 103 completed treatment and had scans at both baseline and week 36.

Higher retatrutide groups lost much more fat mass than lean mass. Lean mass still declined. That supports preferential fat loss, not demonstrated muscle preservation.

The study did not measure muscle strength, physical function, muscle quality or sarcopenia, and its population and 36-week duration remain part of the result. Lilly funded the work, and all seven listed authors reported being Lilly employees and shareholders.

MRI finds smaller muscle volume and less fat inside muscle

SURPASS-3 MRI asked a more specific question in adults with type 2 diabetes. This post-hoc analysis used valid baseline and week-52 scans from 246 people in a randomized, open-label trial.

Across pooled tirzepatide groups, muscle volume fell by 0.64 litres while muscle fat infiltration fell by 0.36 percentage points. The observed volume change did not differ statistically from a prediction based on matched UK Biobank participants. The reduction in fat infiltration was larger than that model predicted.

This is the distinction DXA cannot make: a muscle compartment can become smaller while also containing less fat. Lower infiltration is a muscle-quality signal, not proof of preserved force. And the UK Biobank benchmark was an observational prediction, not a randomized group deliberately matched for the same weight loss. The substudy did not measure strength or physical function.

Direct strength data are reassuring—and too small to settle it

Two newer prospective studies moved beyond scans, but neither supplied a randomized control for the strength question.

In a 12-week proof-of-concept study, ten adults with obesity and type 2 diabetes received semaglutide. Ultrasound showed a smaller vastus lateralis, while knee-extension and handgrip strength did not significantly decline. The authors explicitly said the study lacked the power and duration for firm conclusions.

SEMALEAN followed 115 adults receiving semaglutide 2.4 mg, 106 of whom completed 12 months. Mean lean mass fell by about 3 kg at month 7 and then stabilized. Handgrip strength increased by 4.5 kg at month 12.

Those observations show why function must be measured rather than inferred. They cannot show that semaglutide caused preserved or improved strength: without an untreated randomized comparison, clinical care, practice effects, selection and time remain alternative explanations.

They also do not answer the highest-consequence version of the question. Middle-aged trial cohorts do not establish what happens in older adults, people with frailty, existing sarcopenia or diseases that already accelerate muscle loss.

Meta-analyses inherit the measurement problem

A 2026 meta-analysis of seven placebo-controlled obesity-dose GLP-1 trials found that lean mass became a larger share of body weight while absolute lean mass fell by an estimated 1.74 kg. Heterogeneity for the absolute and relative changes was 98%, meaning the studies did not cluster around one stable effect.

Another 2026 synthesis of 20 randomized trials estimated that lean mass made up 25% to 39% of weight loss across incretin drugs and found no statistically significant proportional difference from lifestyle interventions. Its agents, populations and measurement methods still varied.

These analyses help with proportionality. They cannot manufacture muscle-force or gait outcomes that the underlying trials did not collect.

Original sources

• Wilding et al., STEP 1 main trial
• Wilding et al., STEP 1 DXA exploratory analysis
• Look et al., SURMOUNT-1 DXA substudy
• Coskun et al., retatrutide DXA substudy
• Sattar et al., SURPASS-3 MRI
• Langer et al., direct muscle-size and strength study
• Alissou et al., SEMALEAN
• Laverde et al., obesity-dose GLP-1 meta-analysis
• Eisa and Barood, incretin-versus-lifestyle meta-analysis

Independent editorial summary. The authors, trial investigators, sponsors and journals are not affiliated with LifeScore. Novo Nordisk funded STEP 1. Eli Lilly funded the SURMOUNT-1, SURPASS-3 MRI and retatrutide substudies and employed multiple or all listed authors as disclosed. This article does not provide medical, exercise, nutrition, dosing or treatment advice.