Nearly every entry below is a substudy. No large trial has yet been designed around muscle, so what exists is body-composition scanning attached to obesity trials built to measure weight, plus reviews arguing over what those scans mean. The pattern is broadly similar across drugs. Somewhere between a quarter and 40% of what comes off is lean tissue, and the rest is fat. The single most useful number on this page is from the SURMOUNT-1 scan substudy, where the split was about 75/25 on placebo as well as on the drug. That reframes the question from 'what does this drug do to muscle' to 'what does losing weight do to muscle.' Three things remain unresolved: whether the lean-mass loss affects how people function, how much resistance training and protein offset it, and whether the muscle-sparing drugs now in trials add anything. One limitation runs through everything here. DEXA measures fat-free mass, which includes water and organs, rather than skeletal muscle itself, and almost none of these trials measured strength.
Ranked papers
#1SURMOUNT-1 (body composition substudy)
Look M, Dunn JP, Kushner RF, et al. · Diabetes Obes Metab · 2025
What it measured: Fat, lean tissue and deep abdominal fat on body scan, at 72 weeks
The reference point for this whole question: 160 of SURMOUNT-1's 2,539 participants scanned by DEXA at baseline and week 72. Weight fell 21.3%, fat mass 33.9% and lean mass 10.9% on tirzepatide, against 5.3%, 8.2% and 2.6% on placebo. The finding that matters most is usually left out of the coverage. The split ran roughly three parts fat to one part lean in both groups, and it held in most subgroups by sex, age and how much weight came off. The drug did not shift the composition of weight loss; it produced more of it.
PMID 39996356 ↗NCT04184622 ↗DOI 10.1111/dom.16275 ↗
#2Retatrutide T2D body composition substudy
Coskun T, Wu Q, Schloot NC, et al. · Lancet Diabetes Endocrinol · 2025
What it measured: Fat, lean tissue, abdominal fat and liver fat by scan, at 36 weeks
A scan substudy nested inside retatrutide's phase 2 diabetes program, reading out at 36 weeks. Fat mass fell by up to 26.1% against 4.5% on placebo, and lean tissue made up about the same share of the weight lost as with other obesity treatments. It was published in 2025, after the phase 3 program had started, so it answered the body-composition question rather than clearing the way. Grip strength and physical function were not endpoints, so the functional question is untouched here.
PMID 40609566 ↗DOI 10.1016/S2213-8587(25)00092-0 ↗
#3
Bikou A, Dermiki-Gkana F, Penteris M, et al. · Expert Opinion on Pharmacotherapy · 2024
What it measured: How much of the weight lost on semaglutide was lean tissue, across six trials
The only synthesis built specifically around semaglutide and lean mass: six studies, 1,541 adults with overweight or obesity, summarized narratively rather than pooled statistically. Weight came off predominantly as fat, and lean mass as a share of total body weight went up. But the lean fraction of the weight lost ranged from close to zero in some trials to roughly 40% in others, with the largest reductions appearing in the largest trials. Six heterogeneous studies and no meta-analysis is a weak design. Read it as a map of how widely the estimates scatter; that scatter is the most accurate answer to how much muscle any individual should expect to lose.
PMID 38629387 ↗DOI 10.1080/14656566.2024.2343092 ↗
#4
Lundgren JR, Janus C, Jensen SBK, et al. · N Engl J Med · 2021
What it measured: Weight over a year of maintenance, after everyone had already dieted down
The trial that tests what everyone recommends. 195 adults dieted down over eight weeks first, then spent a year in one of four arms: liraglutide 3.0 mg, a supervised training program, the two together, or neither. The combination lost the most weight (9.5 kg more than placebo, though not significantly more than liraglutide alone) and lowered body-fat percentage by 3.9 points, roughly double what either treatment managed by itself. Only the combination also improved blood sugar, insulin sensitivity and fitness. It is the strongest randomized evidence that training alongside these drugs changes what kind of tissue comes off, not just how much.
PMID 33951361 ↗NCT04122716 ↗DOI 10.1056/NEJMoa2028198 ↗
#5
Jensen SBK, Janus C, Lundgren JR, et al. · EClinicalMedicine · 2024
What it measured: Weight and body composition a year after everything was stopped
The follow-up a year after everything stopped, and the part people rarely hear. A year after the exercise program ended, its group had kept its weight and body composition; after liraglutide alone, weight came back, 6.0 kg more than after exercise. The combination group sat in between and still ended 5.1 kg lighter, with less body fat, than the liraglutide-alone group. Read together with the withdrawal trials elsewhere on this site, it says something uncomfortable: the favorable body composition is a property of continuing, not an achievement banked. Habits built during treatment are what survive it.
PMID 38544798 ↗NCT04122716 ↗DOI 10.1016/j.eclinm.2024.102475 ↗
#6
Locatelli JC, Costa JG, Haynes A, et al. · Diabetes Care · 2024
The review that put a number on the mitigation argument. It puts the lean-mass loss on these drugs at about 10%, or about 6 kg, which it compares to a decade or more of aging, and notes that more than 10 weeks of supervised resistance training can build roughly 3 kg of lean mass and add roughly 25% to strength. Its case is that tailored resistance training should be recommended alongside these drugs to keep muscle while fat comes off.
PMID 38687506 ↗DOI 10.2337/dci23-0100 ↗
#7
Karakasis P, Patoulias D, Fragakis N, et al. · Metabolism · 2025
The network meta-analysis of 22 randomized trials (2,258 participants) of GLP-1 drugs and the GLP-1/GIP dual agonist. Roughly one quarter of the weight people lost was lean mass, and lean mass as a share of body weight did not change. Liraglutide was the only drug that cut weight without significantly reducing lean mass, while tirzepatide 15 mg and semaglutide 2.4 mg removed the most weight and fat but ranked near the bottom for keeping lean mass.
PMID 39719170 ↗DOI 10.1016/j.metabol.2024.156113 ↗
The paper that supplies the vocabulary the rest of the field now uses: adaptive lean-mass loss, proportional to fat loss and physiologically expected, against maladaptive loss that would cost someone function. Its reading of the trial data, with MRI studies added, is that what has been observed so far is the adaptive kind, with less fat inside the muscle. It notes that reported lean-mass losses range from about 15% or less to 40 to 60% of weight lost, that lean mass is not the same thing as muscle, and that older age and more severe disease call for care in choosing patients. Muscle-sparing drug combinations are in development.
PMID 38937282 ↗DOI 10.1111/dom.15728 ↗
#9
Linge J, Birkenfeld AL, Neeland IJ, et al. · Circulation · 2024
A direct argument that the alarm has outrun the evidence, drawing on MRI studies. Muscle volume falls about as much as age, disease and the amount of weight lost would predict, while fat inside the muscle and insulin sensitivity improve, which points to an adaptive response with better muscle quality. Its practical demand is that obesity trials measure muscle properly (its quantity, composition, function, mobility and strength), not a single fat-free-mass number that cannot distinguish muscle from water.
PMID 39401279 ↗DOI 10.1161/CIRCULATIONAHA.124.067676 ↗
#10
Nunn E, Jaiswal N, Gavin M, et al. · Mol Metab · 2024
The preclinical work behind the strategy everyone is watching. In obese mice, semaglutide combined with bimagrumab, an antibody blocking the activin type II receptor, produced more fat loss than semaglutide alone while preserving lean mass. This was in mice, not people. It explains why the combination went into human testing, but it is not evidence that it works in humans; a phase 2 trial in people was published in 2026 and is not on this list.
PMID 38218536 ↗DOI 10.1016/j.molmet.2024.101880 ↗
About this list
Obesity medicine throws up the same handful of questions over and over, and we answer them with ranked lists anchored to the papers themselves. On 2026-10-30 every citation below was put back through PubMed. Where a trial has a registry entry, the card carries that link too, alongside the PubMed one.
Browse our full index of research lists or our long-form research articles.