Lean mass loss on GLP-1s: what the trials measured

In the STEP 1 DEXA substudy, roughly 40% of the total mass participants lost on semaglutide was lean mass. In SURMOUNT-1's body-composition substudy, tirzepatide's split looked closer to 25%. Both numbers are real; neither means what the headlines built on them usually imply. By the end of this article you'll know what the substudies actually measured, what any comparably large weight loss does to lean mass, why the percentage framing misleads, which countermeasures have real evidence, and what a self-tracker can honestly measure at home.

What the substudies actually found

STEP 1 (Wilding et al., NEJM 2021) randomized 1,961 adults with overweight or obesity to semaglutide 2.4 mg weekly or placebo for 68 weeks, with mean weight change of −14.9% versus −2.4%. In a DEXA substudy of about 140 participants, fat mass fell by roughly 8.4 kg and lean body mass by roughly 5.3 kg — lean mass was approximately 40% of the total mass lost. The same substudy also found that the proportion of lean mass relative to total body weight increased: participants ended with a higher lean fraction than they started with.

SURMOUNT-1 (Jastreboff et al., NEJM 2022) randomized 2,539 adults to tirzepatide 5, 10, or 15 mg weekly or placebo for 72 weeks, with mean weight loss reaching −20.9% at the highest dose. Its body-composition substudy (around 160 participants) reported a fat-mass reduction of about 34% versus 8% on placebo, with total mass lost splitting roughly 3:1 fat to lean — lean mass around a quarter of the total.

It is tempting to read 40% versus 25% as tirzepatide sparing muscle better than semaglutide. The substudies were small, the populations and durations differ, and cross-trial comparisons of secondary substudies cannot support that conclusion. No adequately powered head-to-head body-composition trial has settled it.

Large weight loss always includes lean mass

The essential context the headlines drop: losing lean mass alongside fat is what large weight loss does, by any method. Heymsfield et al. (Obesity Reviews, 2014) critically reviewed the long-standing rule that about one-fourth of diet-induced weight loss is fat-free mass, finding the true proportion varies with the size of the deficit, protein intake, training status, sex, and age — but typically lands in the 20–30% range.

Against that baseline, SURMOUNT-1's ~25% sits exactly where conventional caloric restriction lands, and STEP 1's ~40% sits at the high end of the range. The drugs are not uniquely catabolic; a large, sustained energy deficit is doing what large energy deficits do. What is new is scale. Mean reductions of 15–21% of body weight were previously bariatric-surgery territory, so the absolute lean losses are larger than diet studies historically produced — which is why the question deserves attention even though the mechanism isn't novel.

Why "40% of the loss was lean" misleads

Four reasons the headline number tells you less than it seems to.

DEXA "lean mass" is not muscle. Lean soft tissue on a DEXA report is everything that isn't fat or bone mineral: water, glycogen, organ tissue, gut contents, and skeletal muscle. The first weeks of any energy deficit drain glycogen and its bound water — several pounds that are booked as "lean" but are not contractile tissue and return as quickly as they left.

A scary ratio can coexist with improving composition. Lose 8.4 kg of fat and 5.3 kg of lean, and the lean fraction of your body goes up — which is precisely what STEP 1's substudy reported. "40% of the loss was lean" and "participants ended proportionally leaner" describe the same data.

Absolute numbers need context. Five kilograms of lean mass means something different to a 40-year-old lifter at 110 kg than to a sedentary 70-year-old at 80 kg. The clinical concern — sarcopenia, frailty, falls — concentrates in older adults and in repeat weight-loss cycles, not in the average headline reader.

Function was not measured. Neither substudy reported strength or physical-performance outcomes. Composition is a proxy for what people actually care about, and an imperfect one; a kilogram of DEXA-lean lost tells you nothing by itself about whether you can still do what you could before.

Countermeasures with real evidence

An honesty note first: there is not yet a large published randomized trial of resistance training or protein supplementation specifically in GLP-1 users. The evidence comes from the general weight-loss literature. It is, however, consistent and strong.

Resistance training. Villareal et al. (NEJM 2017) put obese older adults through a ~10% diet-induced weight loss with aerobic, resistance, or combined exercise: the groups doing resistance work lost markedly less lean mass and bone density than the aerobic-only group, while physical-performance scores improved most with combined training. Mechanistically unsurprising — the stimulus to keep muscle is using it under load — but it's the rare mitigation with hard trial data behind it.

Protein intake. Wycherley et al. (AJCN 2012), a meta-analysis of 24 energy-restriction trials, found higher-protein diets retained significantly more fat-free mass than standard-protein diets at matched calorie deficits. Longland et al. (AJCN 2016) sharpened the point under harder conditions: in a 40% energy deficit with six-day-a-week training, the group eating 2.4 g/kg/day of protein gained lean mass while losing more fat than the 1.2 g/kg/day group. Those are the trial facts; both protein targets and deficit sizes in these studies were set against body weight and energy expenditure, which is the arithmetic the site's TDEE calculator covers.

Appetite suppression is the practical catch. A compound that halves your appetite makes hitting any given protein intake mechanically harder — the mitigation with the best evidence is the one GLP-1 pharmacology works against. That tension, not a novel catabolic mechanism, is arguably the real muscle story with this class.

The pipeline: combinations designed to spare muscle

The question is well enough understood that drug developers are targeting it directly. Bimagrumab — an investigational activin type II receptor antibody, not approved for any indication — showed an unusual profile in a phase 2 trial in adults with type 2 diabetes and obesity (Heymsfield et al., JAMA Network Open, 2021): fat mass fell by roughly 20% while lean mass increased by around 4% over 48 weeks.

That profile made it an obvious combination candidate. The BELIEVE phase 2b trial (started by Versanis, continued by Eli Lilly after its acquisition) tested bimagrumab alone and combined with semaglutide; topline results reported in 2025 indicated that a greater share of weight lost with the combination was fat. Peer-reviewed publication was still pending at this writing, so the details deserve the usual caution. Other muscle-preservation agents are in earlier-stage trials alongside GLP-1s, but bimagrumab is the furthest along of the publicly reported combinations.

What a self-tracker can actually measure

Every instrument available outside a research center measures something different, with different noise. Trend honesty means knowing which is which.

Scale weight measures total mass only, with ±1–2 kg of day-to-day noise from water, glycogen, sodium, and gut contents. It says nothing about composition, but it's the cheapest high-frequency signal there is — useful as a weekly trend, meaningless as a daily verdict.

Impedance (BIA) smart scales estimate body fat from electrical conductance, which makes them hydration meters as much as fat meters. Absolute readings can be off by several percentage points, and readings don't transfer across devices. A trend taken under fixed conditions — same scale, morning, post-bathroom, pre-coffee — carries some signal; everything else is noise wearing a decimal point.

DEXA is the strongest consumer-accessible composition split, and it's what both trial substudies used. It still isn't a muscle meter (see above), and scan-to-scan precision means lean-mass changes under roughly a kilogram are within noise. Same facility, same machine, similar hydration — or the comparison is corrupted before it starts.

A strength log is the cheapest functional proxy, and it measures the thing the trial substudies didn't: whether performance is holding while mass falls. Load, reps, and grip don't care what the impedance scale thinks.

The common thread: pick one instrument per quantity, keep conditions fixed, and judge trends rather than single readings. Switching tools mid-experiment manufactures phantom changes — and with lean mass, the phantom changes are usually bigger than the real ones.

Sources

• Wilding JPH et al. Once-weekly semaglutide in adults with overweight or obesity (STEP 1, including the DEXA substudy). New England Journal of Medicine, 2021.
• Jastreboff AM et al. Tirzepatide once weekly for the treatment of obesity (SURMOUNT-1, including the body-composition substudy). New England Journal of Medicine, 2022.
• Heymsfield SB et al. Weight loss composition is one-fourth fat-free mass: a critical review and critique of this widely cited rule. Obesity Reviews, 2014.
• Wycherley TP et al. Effects of energy-restricted high-protein, low-fat compared with standard-protein, low-fat diets: a meta-analysis of randomized controlled trials. American Journal of Clinical Nutrition, 2012.
• Longland TM et al. Higher compared with lower dietary protein during an energy deficit combined with intense exercise promotes greater lean mass gain and fat mass loss. American Journal of Clinical Nutrition, 2016.
• Villareal DT et al. Aerobic or resistance exercise, or both, in dieting obese older adults. New England Journal of Medicine, 2017.
• Heymsfield SB et al. Effect of bimagrumab vs placebo on body fat mass among adults with type 2 diabetes and obesity: a phase 2 randomized clinical trial. JAMA Network Open, 2021.
• Eli Lilly. BELIEVE phase 2b topline results (bimagrumab alone and with semaglutide), 2025.