Fat versus lean: what the STEP 1 and SURMOUNT-1 DXA substudies actually measured

Both pivotal trials scanned a subset of participants with dual-energy X-ray absorptiometry at baseline and at the end. The samples are small, the analyses were exploratory, and the numbers are the best direct evidence there is about what the weight lost on these drugs is made of.

STEP 1: semaglutide 2.4 mg, 68 weeks

140 participants (95 semaglutide, 45 placebo) had DXA scans. Their mean baseline weight was 98.4 kg and BMI 34.8; 76 percent were women. In the semaglutide group, body weight fell 15.0 percent (versus 3.6 percent on placebo), total fat mass fell 19.3 percent and visceral fat mass 27.4 percent. Total lean body mass fell 9.7 percent in absolute terms, but because fat fell faster, the proportion of the body that was lean rose by 3.0 percentage points. The lean-to-fat ratio improved most in those who lost 15 percent or more of body weight (Wilding 2021, J Endocr Soc abstract A16; PubMed 33567185 for the main trial).

Applying those percentages to the substudy's baseline composition (43.4 percent fat, 53.9 percent lean of 98.4 kg) gives roughly 8.2 kg of fat and 5.1 kg of lean tissue lost, so lean tissue was about 38 percent of the mass lost by DXA. That derivation is ours; the paper reports the percentages.

SURMOUNT-1: tirzepatide, 72 weeks

160 participants (124 on pooled tirzepatide doses, 36 placebo) had scans. Mean baseline weight 102.5 kg, BMI 38.0, 73 percent women. Body weight fell 21.3 percent with tirzepatide versus 5.3 percent with placebo; fat mass fell 33.9 versus 8.2 percent; lean mass fell 10.9 versus 2.6 percent. Of the body weight lost, approximately 75 percent was fat mass and 25 percent lean mass, and that split was the same for tirzepatide and for placebo, and held across subgroups by sex, age and amount lost (PubMed 39996356).

That last point deserves emphasis. The placebo group, losing weight through diet and activity counselling alone, lost it in the same 75:25 proportion. On this evidence the drug does not change the composition of weight loss; it changes the amount.

How that compares

Neeland and colleagues reviewed the literature on lean mass with GLP-1-based therapies in 2024 and found wide heterogeneity: in some studies lean mass reductions accounted for 40 to 60 percent of total weight lost, in others 15 percent or less (PubMed 38937282). Method matters (DXA lean mass is not muscle; MRI measures muscle directly and tends to show smaller losses), as do the population's age, starting muscle mass, protein intake and training. The 25 to 38 percent range in the two pivotal substudies sits in the middle of the reported spread and in line with what is typically seen with diet-induced loss.

What "lean mass" includes

DXA divides the body into bone mineral, fat and everything else, which it calls lean soft tissue. Lean soft tissue is skeletal muscle, but also organ tissue, skin, connective tissue, body water and the protein scaffolding inside fat tissue. When fat mass falls by a third, the non-fat components of fat tissue fall with it and are counted as lean. When glycogen stores drop early in a deficit, the water bound to them leaves and is counted as lean. Neither is muscle. The muscle component of a DXA lean loss is real but smaller than the headline figure.

None of this makes lean loss trivial. In older adults, in people who start with low muscle mass, and at the higher end of the reported range, losing muscle costs strength, function and resting energy expenditure, and it makes regain more likely to arrive as fat. It makes the question answerable: how much of my loss is muscle, and what changes the share?

What changes the share

Two things, in every population where it has been tested, and the trials are described in detail on their own pages:

• Resistance training. Villareal's trial in older adults with obesity found lean mass fell less in the groups that lifted than in the group that did aerobic exercise only, at the same weight loss (PubMed 28514618). See resistance training basics and exercise plus GLP-1 trials.
• Protein. Longland's trial showed lean mass rising during a steep deficit when protein was high and training was intense (PubMed 26817506); Morton's meta-analysis puts the useful ceiling at about 1.6 g/kg/day (PubMed 28698222). See protein targets.

The Neeland review lists both as the primary mitigation strategies, alongside emerging pharmacological approaches that are not yet approved (PubMed 38937282). The 2025 joint advisory turns them into the three-times-weekly strength training and 1.2 to 1.6 g/kg/day protein that this site's calculator uses.

Measuring it yourself

A DXA scan before or early in therapy and another six to twelve months later answers the question for you rather than for a trial population. Waist circumference monthly and a lift log every session are the free proxies. See body composition measures. If you are on a compounded product, note that it is not FDA approved and not interchangeable with the brand products these substudies used; the physiology of weight loss composition is not product-specific, but the numbers above were measured on brand products at label doses.