What the scale hides: lean mass in GLP-1 weight loss

The headline number in a GLP-1 trial is total weight lost. It is the wrong number to design a programme around.

What the trials measured

In STEP 1, 1,961 adults on semaglutide 2.4mg lost a mean 14.9% of body weight over 68 weeks. In SURMOUNT-1, 2,539 adults on tirzepatide lost a mean 20.9% over 72 weeks.

The body composition sub-analyses are where it gets interesting, and where the two drugs separate. In STEP 1, lean mass fell 6.92 kg against 15.3 kg of total weight, which is roughly 45% of the loss. In SURMOUNT-1, lean mass fell 5.67 kg against 22.1 kg, roughly 26%. Across the wider literature the figure runs between about 26% and 40% depending on the trial, the population, and how body composition was measured.

Two things follow. The proportion is large in every case, and it is not the same number for every drug, so “a quarter to a third” is a summary that flatters semaglutide.

That proportion is also not a GLP-1 artefact. It broadly mirrors what happens in diet-alone weight loss. The mechanism is unremarkable: in a caloric deficit the body mobilises stored energy from several tissues at once, and it does not preferentially spare muscle. Where protein intake falls short, it breaks down muscle to reach the amino acids.

Why the ratio matters more than the total

Muscle is metabolically active tissue. Losing it lowers resting metabolic rate, which makes maintenance harder at precisely the point where medication is reduced or stopped. Layered on top is the ordinary trajectory of ageing: most adults lose 3% to 5% of muscle mass per decade after 30, and the rate climbs later in life. For an older cohort, the same percentage carries meaningfully more clinical weight.

Stated plainly: losing 40 pounds of mostly fat while holding lean mass is a better outcome than losing 50 and giving back a third of it in muscle.

What actually moves it

Two interventions have support, and both are unglamorous.

Resistance training. Two to three sessions a week, 30 to 45 minutes, compound movements across the major muscle groups, with progressive overload. Bodyweight and bands are sufficient to start; equipment is not the variable. The evidence base for this is covered in Field Note 002, including where it runs out.

Protein intake. A 2024 systematic review and meta-analysis found that intake above 1.3g per kilogram per day was associated with preserved or increased muscle mass in adults with overweight or obesity losing weight, while intake below 1.0g/kg/day was associated with decline. The AACE 2025 consensus statement recommends at or above 1.2g/kg/day for most individuals, roughly one and a half times the 0.8g/kg RDA. This is the harder of the two to deliver, because appetite suppression is the mechanism of action. Reduced intake and elevated protein requirement pull in opposite directions, and closing that gap takes structure rather than instruction.

Where the evidence runs out

SURMOUNT-5 compared tirzepatide against semaglutide head to head, but did not report detailed body composition analysis. The STEP 1 and SURMOUNT-1 figures above come from separate trials in different populations, so the apparent gap between the two drugs is suggestive rather than a controlled comparison. There is a plausible hypothesis that tirzepatide’s GIP receptor activity confers some lean mass advantage. It has not been tested in a rigorous head-to-head body composition trial, and we do not present it as established.

Note also that the trials generating these figures were designed to test medication efficacy, not to optimise muscle preservation. They describe what happens under ordinary conditions. They do not establish the ceiling on what a well-constructed adjunct programme could achieve. That, from our perspective, is the more interesting open question.

What we do with it

Body composition, not scale weight, is the measurement that matters. Waist circumference is a practical proxy; DEXA is the reference standard where it is available. Any programme pairing GLP-1 therapy with serious lean-mass work should be measuring the ratio rather than the total, and should be prepared to publish it.