The self-quantified, data-to-action human.
The muscle question is the one every endurance athlete asks about GLP-1 before they ask anything else, and it is the one I can answer least cleanly. I have been on some version of this protocol since February 2026. The dosing history, the drug switches, the side-effect record and the week-by-week weights all live on the GLP-1 hub, and I am not going to restate them here. This page is about one narrower thing: what my body-composition log says about lean mass, and how much of that I actually believe.
The measurement problem comes first
There is no DEXA scanner within convenient reach, so body composition here is weekly circumference measurements — weight, waist, belly button, chest, glutes, both legs, both arms — pushed through a linear regression model for body fat and lean mass. That is a worse instrument than a scan and I want that stated before any number appears. What it is good at is trend. Taken at the same time of day, in the same order, with the same tape, it moves in the right direction when something real is happening. What it is bad at is telling you which tissue moved.
That distinction turned out to matter more than I expected, because the two signals disagreed.
Where the tape said one thing and the model said another
Across the first five-week block the circumference pattern looked like a textbook good outcome. The loss concentrated in the trunk: waist down 2.0 cm, belly button down 3.5 cm. Chest was flat. Glutes, legs and arms all moved less than 2 cm, and most of what they did move came back as weight rebounded afterwards, which reads far more like glycogen and water than like leg tissue leaving. If you judged muscle preservation by the tape alone, you would call that a win and move on.
The model underneath disagreed. Estimated lean mass went from 83.3 kg to 78.9 kg over the same window. Roughly 4 kg. I was doing everything the literature tells you to do — full-body resistance training throughout, high protein, training volume held, not a single session missed to the medication — and about 4 kg of estimated lean mass came off anyway.
I have stared at that pair of numbers for months. The honest reading is that I do not know which one is closer to the truth. A circumference model partially infers lean mass from weight, so a fast drop in weight will drag the lean estimate down whether or not any muscle actually left. Equally, a limb can lose real cross-sectional muscle and gain enough intramuscular water to hold the tape steady. Both stories fit. That is what an ambiguous dataset looks like, and dressing it up as a finding would be the exact thing this pillar exists to avoid.
What I ran alongside it
The protocol was not sophisticated, which is probably why it was sustainable. Resistance work stayed in the week the entire time, treated as non-negotiable rather than as something to drop when the endurance block got heavy. Protein went up and stayed up. Injections moved to Friday evening, right before bed, which killed the side effects almost entirely and meant no training session ever landed inside a nausea window. Endurance volume did not get cut to accommodate the drug.
The part people underestimate is fueling. GLP-1 slows gastric emptying, so protein does not just need to be eaten, it needs to be eaten early enough and in small enough portions to actually land. On a suppressed appetite, hitting a protein target is a logistics problem before it is a nutrition problem. Days where I got it right were days I had planned the meals before I was hungry, which on this drug is most of the day.
The variance nobody warns you about
The clearest thing in the whole food log is not a deficit. It is spread. A stretch of eight logged days in late summer averaged almost exactly my calorie target, and the same days ran from about 1,340 kcal to about 5,360 kcal, with both extremes inside a single five-day window. That is a restrict-then-correct cycle. For fat loss it roughly cancels out. For building or holding muscle it is close to the worst possible shape, because the protein on a 1,340-kcal day is not there, and the surplus on the 5,360-kcal day arrives as something I did not plan to eat rather than as anything useful. If I had one intervention to run again, it would be flattening that curve rather than touching the dose. The dose-escalation experiment is written up in the kill list, and it did not work.
What I would tell another endurance athlete
Lifting and protein are not optional insurance on this drug, they are the protocol. They also did not fully protect me, at least not by the measure I have. The tape held; the model did not. If you have access to a DEXA, get a baseline before you start, because the version of this question I can answer is much weaker than the version you could.
The one thing I would not do is read a stable arm measurement as proof that nothing left. That is exactly the mistake I nearly made. For the recovery-side version of the same lesson — a leg that measured fine and could not do anything — see Rebuilding a Femur on Data. Back to the Biohacking hub.
Ready to Train Smarter?
Get personalized training zones, race predictions, and performance insights with our free calculators.
