How much of the weight lost on a GLP-1 drug is lean mass
In the dual-energy X-ray absorptiometry substudy of the 68-week phase 3 semaglutide obesity trial, fat-free mass accounted for something on the order of 40 percent of the total weight lost. In the body composition substudy of the 72-week tirzepatide obesity trial, the split was closer to three parts fat to one part lean, so roughly a quarter. Those are the two best-characterised numbers in the field, and they are not the same number, which is itself the first useful finding.
Both substudies also reported the fact that gets lost in headlines. The proportion of the body that was fat went down and the proportion that was lean went up. Participants finished lighter and, as a fraction of their own body, leaner than they started. That is the arithmetic signature of a fat-predominant loss, not a wasting syndrome.
The number that should worry a reader is not the percentage but the absolute quantity. A person who loses 3 kg on a diet and gives up a quarter of it as lean tissue has lost under a kilogram of fat-free mass. A person who loses 20 kg on tirzepatide and gives up the same quarter has lost 5 kg. The ratio is reassuring and the tonnage is not, and both statements are true at once.
Fat-free mass is not the same thing as muscle
A scanner divides soft tissue into fat and not-fat. The not-fat compartment is skeletal muscle plus liver, kidneys, gut, heart, skin, tendon, blood, and the extracellular water sitting in all of it. In an adult, skeletal muscle is only about half of total fat-free mass. So a report that fat-free mass fell by 5 kg is compatible with anything from a serious loss of contractile tissue to almost none of it.
Some of the earliest and largest lean decline on any weight loss intervention is water and glycogen. Each gram of stored glycogen holds roughly three grams of water with it, and glycogen stores fall sharply in the first weeks of an energy deficit. Plasma volume contracts too, and gastrointestinal contents drop when someone eats substantially less, which is a live consideration for drugs that slow gastric emptying. None of that is muscle, and all of it reads as lean mass loss on a scan taken at week 12.
Obesity itself inflates the lean compartment. Larger bodies carry larger organs, more connective tissue threaded through adipose depots, expanded blood volume and genuinely more skeletal muscle, because moving 130 kg around is a training stimulus. Shedding part of that scaffolding as the load comes off is the expected physiology of getting smaller, not a side effect peculiar to incretins.
What the incretin trials actually measured
The phase 3 semaglutide trial randomised 1,961 adults with obesity and without diabetes to weekly semaglutide 2.4 mg or placebo for 68 weeks, and reported mean weight loss of about 15 percent against about 2 percent on placebo. Body composition was not the point of the trial. It was assessed by scan in a subset of roughly 140 participants at selected sites, as an exploratory measure, with no power calculation behind it.
The phase 3 tirzepatide obesity trial randomised 2,539 adults for 72 weeks and reported mean weight loss of roughly a fifth of body weight at the highest dose against about 3 percent on placebo. Its body composition substudy, again around 160 people, found fat mass falling by roughly a third while lean mass fell by around a tenth, which is where the three-to-one figure comes from.
Retatrutide, the triple agonist at the GIP, GLP-1 and glucagon receptors, produced about 24 percent mean weight loss at its highest studied dose over 48 weeks in phase 2. It is not approved by any regulator, its phase 3 obesity programme is still running, and its published body composition reporting comes from small imaging subsets. Any claim that it spares or destroys lean tissue more than the incretins already on the market is, at the time of writing, an extrapolation rather than a finding.
It is worth being blunt about what these substudies are. They are opt-in subgroups, small enough that a handful of dropouts moves the mean, measured at two timepoints with a technique whose precision for one person is worse than the effect reported. They are the best evidence available and still exploratory endpoints in trials designed to answer a different question.
The right comparison is matched weight loss, not zero
Every method of losing weight loses fat-free mass. Gilbert Forbes established decades ago that the fat-free fraction of weight lost is predictable, and that it is larger in leaner people and smaller in people with a lot of fat to give up. Later pooled analyses across diets, very-low-calorie regimens and surgery landed repeatedly in the same neighbourhood: roughly 20 to 30 percent of what comes off is fat-free mass.
Bariatric surgery sits in that band too, often at the upper end, because the loss is fast and large. Nobody describes sleeve gastrectomy as a muscle-wasting operation, and the long-term outcome data on surgery are the strongest in the field. That is a useful calibration for how much weight to put on a lean-mass percentage in isolation.
So the question that would actually settle the argument is narrow: for two people who lose the same 18 kg, one on semaglutide and one on a supervised diet, does the incretin group give up more fat-free mass per kilogram lost? A trial designed to answer that, with matched weight loss as the entry condition rather than the outcome, has not been run. Cross-trial comparison cannot substitute, because the populations, timescales and scanners differ.
Scans disagree: DXA, MRI and D3-creatine
Dual-energy X-ray absorptiometry is cheap, fast and reproducible enough for group means, which is why it dominates these substudies. Its weakness is that it infers soft tissue composition from X-ray attenuation and assumes a fixed hydration of the lean compartment. When hydration shifts, and it always shifts during rapid weight loss, part of what it reports as lean change is fluid.
Magnetic resonance imaging and computed tomography measure muscle volume or cross-sectional area directly, and can separate muscle from the fat infiltrating it. That distinction matters because fat infiltration of muscle is itself associated with weakness, and imaging in metabolic weight loss has suggested that intramuscular fat falls alongside muscle volume. Muscle can get smaller and denser at the same time, and a scan reporting only lean kilograms cannot see that.
Deuterated creatine dilution estimates whole-body skeletal muscle mass from tracer dilution rather than from a density assumption, and in ageing cohorts it tracks strength and physical function better than lean mass by scan does. It has barely been used in the incretin trials. That absence, more than any number that has been published, is why confident statements about muscle specifically are premature.
Mass went down and function generally did not
The trial that speaks most directly to function enrolled people with obesity-related heart failure with preserved ejection fraction and gave them semaglutide. Symptom and physical limitation scores improved substantially and six-minute walk distance increased. Whatever lean mass those participants lost, they walked further at the end than at the start.
The cardiovascular outcomes trial in more than 17,000 adults with established cardiovascular disease and overweight or obesity without diabetes reported roughly a 20 percent reduction in major adverse cardiovascular events over several years of follow-up. Hard outcomes moved in the right direction. A drug that was meaningfully degrading skeletal muscle across a population that size would be an odd fit with that result.
The caveats are real and specific. These trials enrolled mostly middle-aged adults with functional reserve to spare, they rarely measured strength directly rather than walking or symptom scores, and adults over 75, frail patients and people with existing sarcopenia were thinly represented or excluded. An absence of a functional signal in a robust population is not evidence of safety in a fragile one.
What resistance training and protein do, and do not, fix
Outside the incretin literature the evidence is reasonably consistent. Adding resistance training to an energy deficit preserves a meaningful share of fat-free mass that would otherwise be lost, and raising protein intake within the range commonly studied in weight loss trials attenuates it further. The randomised trial in older adults with obesity that compared diet, exercise, diet plus exercise and control found that the combination preserved lean mass and improved physical function more than dieting alone.
Two things make that harder to transplant onto an incretin. First, energy intake falls far enough that hitting a protein target becomes a genuine logistical problem rather than a preference, and nausea and early satiety tend to push intake away from dense protein sources specifically. Second, older muscle is anabolically resistant, needing a larger protein stimulus to mount the same synthetic response, and older adults are exactly the group with most to lose.
The direct evidence in people taking these drugs is thin: small trials, short follow-up, scan endpoints rather than function, and inconsistent supervision of the training itself. The mechanistic case is strong and the intervention carries little downside, but the adequately powered trial showing that resistance training plus higher protein changes strength or physical function on semaglutide or tirzepatide has not been run.
Drugs designed to keep the muscle
Bimagrumab, an antibody against the activin type II receptor, is the most interesting data point. In a 48-week phase 2 trial in adults with type 2 diabetes and obesity it reduced fat mass by around a fifth while lean mass increased slightly, which is a dissociation of the two compartments that diet cannot achieve. Total weight change was modest, and no functional benefit was demonstrated.
The obvious combination has been tested. A phase 2b trial of bimagrumab with semaglutide reported that a substantially greater share of total weight lost was fat compared with semaglutide alone. Separately, the selective androgen receptor modulator enobosarm and the myostatin antibody trevogrumab have both entered phase 2 combination trials with semaglutide, each reporting body composition endpoints.
Every one of these programmes is phase 2, every headline result is a surrogate endpoint measured by scan, and none has shown that a person is stronger, walks further, falls less or lives longer. Regulators have signalled that body composition alone will not carry an approval here. A drug that improves a scan while leaving function unchanged is a documented failure mode in muscle pharmacology, not a hypothetical one.
Who is plausibly at risk
The risk is not uniformly distributed, and pooled trial means hide that. The clearest cases are older adults who already meet criteria for sarcopenic obesity, in whom muscle mass and strength are low before treatment starts; people entering treatment at the lower end of the obesity range, where the fat-free fraction of loss is predictably higher; and anyone losing weight very fast.
Weight cycling deserves more attention than it gets. In the withdrawal extension of the phase 3 semaglutide trial, participants regained about two-thirds of the lost weight within a year of stopping. Regained weight is disproportionately fat, so the arithmetic of repeated loss-and-regain cycles points toward a body that is progressively fatter at the same weight. That is a mechanistically plausible ratchet, not a demonstrated one, and it has never been tracked across multiple cycles with incretins.
Bone is the loose end. Bone mineral density falls with large weight loss by almost any method, the incretin substudies are far too small and too short to resolve fracture risk, and the ageing population now being treated is the population in which fractures matter most. That gap is not an argument against the drugs; it is an argument that the question has not been asked properly yet.
How to read the next body composition headline
Five questions dispose of most of them. Which compartment was measured, fat-free mass or skeletal muscle? Measured how, by scan or imaging or tracer? Over what interval, and how much of the early change could be water? Against what comparator? And was that comparator matched for weight lost, or is it a placebo group that barely lost anything, which guarantees a dramatic-looking difference for trivial reasons?
Watch for the percentage-versus-kilogram switch, which cuts both ways. Advocates quote the falling body fat percentage and the rising lean percentage. Critics quote the absolute kilograms of fat-free mass. Both are drawn from the same dataset and neither is a lie. An honest account gives both, and then says whether anyone measured what the participants could actually do.