What the Controlled Trials Actually Show
The GLP-1 muscle question cannot be settled by a single headline or a single scan. The clearest evidence in the supplied material comes from randomized trials of obesity-approved doses, while a broader review adds studies from people with and without type 2 diabetes, different formulations, and different follow-up periods. Reading those two layers separately shows what is well supported: GLP-1 treatment usually removes more fat than lean tissue, but the size and meaning of lean-mass change vary enough that no pooled number is a personal forecast.
The narrowest answer: obesity-dose randomized evidence
The first review deliberately asked a focused question. It included adults with obesity, with or without diabetes, who received a GLP-1 receptor agonist at a dose approved for obesity management and were compared with placebo. The eligible regimens were liraglutide at 3 milligrams daily, semaglutide at 2.4 milligrams weekly, and tirzepatide at 5, 10, or 15 milligrams weekly. The review found seven randomized controlled trials involving 821 participants; six contributed to the meta-analysis.
That design deserves special weight because randomization and a placebo comparison make it easier to attribute an average difference to the treatment rather than to motivation, background care, or the natural course of weight change. The included trials were judged to have low overall risk of bias, although two raised concerns about selective outcome reporting. The lean-mass outcomes received high certainty in the review’s GRADE assessment. This does not make the evidence perfect: the sample was small for a class-wide question, the participants were mostly relatively young and non-diabetic, and the review could not reliably assess publication bias because fewer than ten randomized studies were available.
The most important result is that three different lean-mass outcomes moved in different directions. Lean mass as a proportion of total body weight increased by 1.81 percentage points, with a 95% confidence interval from 1.10 to 2.52 percentage points. In ordinary language, lean tissue made up a larger share of the body that remained after treatment. The result was statistically consistent across the included studies, with low heterogeneity: I² was 7%.
The absolute result was less reassuring but not necessarily alarming. Absolute lean mass decreased by an average of 1.74 kilograms compared with placebo, with a 95% confidence interval from 3.04 kilograms lower to 0.45 kilograms lower. The percentage change from baseline also decreased, by 3.06 percentage points on average, with a 95% confidence interval from 5.10 percentage points lower to 1.02 percentage points lower. These figures describe a smaller measured lean compartment, not automatically a loss of the same amount of contractile muscle. They also do not contradict the higher lean-mass proportion: fat mass fell faster than the broad lean compartment.
The difference between consistency and magnitude is crucial. The I² statistic was 98% for both the absolute and percentage lean-mass outcomes. I² estimates how much variation among study results is attributable to differences between studies rather than chance. An I² of 98% does not erase the pooled direction, but it warns that the average conceals major differences in setting, intervention, or measurement. The same number can therefore support two statements at once: the randomized evidence shows an average reduction in measured lean mass, and the pooled reduction is not a reliable prediction for every person receiving treatment.
Subgroups make that variation visible. In trials of liraglutide lasting less than 52 weeks, the absolute difference in lean mass was about 0.41 kilograms lower than the comparator. In the longer liraglutide subgroup, the difference was about 2.65 kilograms lower, although that estimate came from limited data and had moderate heterogeneity.