How much of GLP-1 weight loss is actually lean mass, and where does the "25%" figure come from?

The claim that "about a quarter of what you lose on these drugs is muscle" gets repeated everywhere and I have never once seen it sourced. I am 14 weeks into tirzepatide, down 11.2 kg from 104.6 kg, and I would like to know what fraction of that is realistically lean tissue rather than guessing.

Three specific things I cannot resolve on my own:

• Does the 25% figure come from an actual published body-composition substudy, or is it someone's back-of-envelope from total weight curves?
• Is "lean mass" in those papers the same thing as muscle? My scale calls everything that is not fat "muscle" and I assume that is wrong.
• Is the fraction different on a GLP-1 than on plain caloric restriction? The way it is talked about implies the drug specifically spares muscle, and I cannot find the head-to-head that would show that.

I do not have a baseline DEXA, which I now regret. I have a grip dynamometer reading from a physio appointment last year (right hand 51 kg) and nothing else objective. If the honest answer is "you cannot know retrospectively", I will take that, but I would still like to know what the trial data says the expected split is so I know whether what I am seeing in the mirror is normal or a problem.

Separate point from the accepted answer, and the one that trips up most people reading their own scans: "lean mass" on a DEXA report is lean soft tissue. It is everything that is not fat and not bone mineral. That includes skeletal muscle, but also organs, connective tissue, gut contents to some extent, blood volume, intracellular and extracellular water, and glycogen.

The water and glycogen component is not a rounding error during rapid loss. Work it through:

Trained-ish adult glycogen stores:
  liver          ~ 80-110 g
  skeletal muscle ~ 350-500 g
  total          ~ 430-610 g

Each gram of stored glycogen binds ~3 g of water:
  500 g glycogen x 3 g water = 1500 g water
  total glycogen + bound water = 500 + 1500 = 2000 g = 2.0 kg

So a person who goes from eating 250 g of carbohydrate a day to eating 90 g, which is what happens by accident when appetite falls off a cliff, can drop roughly 1.5-2.0 kg that DEXA will happily score as lost lean soft tissue. Zero contractile protein has been lost. Add sodium and aldosterone shifts on top and the first-month "lean loss" on a scan is substantially fluid.

The practical consequences:

• A DEXA taken in the first 8-12 weeks of aggressive loss overstates lean tissue loss. Do not panic-interpret it.
• The same distortion works in reverse when you stabilise. People who add carbohydrate back at maintenance "gain lean mass" on the next scan without training harder. That is refilling, not hypertrophy.
• Scan under standardised conditions or the noise swamps the signal: same time of day, fasted, no training in the previous 24-48 hours, similar carbohydrate intake for the preceding 2-3 days, well hydrated but not water-loaded.

None of this means lean loss is imaginary. It means the first scan-to-scan delta is the least trustworthy one you will ever take, and that the signal only becomes clean once your intake and training have been stable for a couple of months.

On your third question, whether the fraction differs from plain caloric restriction: the cleanest evidence that the drug is not the variable comes from the liraglutide-plus-exercise randomised trial, which ran four arms after a low-calorie induction phase: placebo, exercise alone, liraglutide alone, and both combined [2].

The headline everyone quotes is that the combination produced the largest fat loss. The part relevant to you is the body-composition breakdown: the arms that included supervised exercise held lean mass, and the drug-alone arm did not. The incretin did not confer protection by itself. The training did.

That is consistent with the broader literature on energy restriction going back decades: the two levers that reliably move the fat-to-lean split are (1) a mechanical loading stimulus and (2) adequate protein. Rate of loss is a weak third lever. The pharmacology is orthogonal to all three; what it changes is your ability to actually run the deficit.

Which reframes your question usefully. You are asking "how much muscle has the drug cost me". The better question is "what fraction of my loss is lean given that I am running a deficit of roughly X, eating Y grams of protein, and lifting Z times per week". Fill those in and the expected lean fraction ranges from near zero (well-fed, lifting, moderate deficit, high baseline fat mass) to 35-40% (very low intake, no loading, already lean, older). Same drug in both cases.

Practical answer to "I have no baseline": you cannot recover the counterfactual, so stop trying and start building forward data.

Get a DEXA now. It will not tell you what you lost, but it becomes the floor against which the next 6-12 months are measured, and that is the comparison that actually matters for whether you have a problem.

In the meantime you do have one useful retrospective anchor: that 51 kg grip reading. Re-test it on the same style of dynamometer, same hand, same posture, same instruction, best of three. Grip is a crude but genuinely predictive proxy for whole-body strength, and it is one of the first things to sag when someone is under-eating protein and under-recovering. If you are within a couple of kilograms of 51, whatever you lost has not meaningfully compromised force production. If you are at 40, that is worth acting on regardless of what any scan says.

Add two zero-cost functional markers you can repeat monthly: a timed five-repetition chair stand, and the load at which a familiar compound lift feels like a hard set of eight. Function is what lean mass is for. Track function.

The other retrospective anchor worth constructing, and it costs nothing: dig out any photograph of yourself from before week zero in which your arms and shoulders are visible, ideally in similar light. Photographs are terrible for quantifying fat and surprisingly informative about limb girth over a 10 kg change. Combine that with a tape measure at three fixed sites, mid-upper arm, mid-thigh and calf, taken at the same landmarks monthly from now on. Circumference is a crude proxy that conflates muscle and fat, but a thigh that is holding its circumference while your waist falls 8 cm is not a thigh that is being dismantled.

None of these substitutes for a scan. Together they will tell you within three months whether anything is actually going wrong, which is the only question that requires an answer.