GLP-1 Lean Mass Loss by Drug: 2026 Trial Data
The RX Index is an independent research and reference resource for primary-sourced prescription-medication data. This page is educational, has not been reviewed by a clinician, and contains no advertising, affiliate links, provider placements, lead forms or product recommendations.
GLP-1 lean mass loss by drug: the verified comparison
GLP-1 lean mass loss by drug cannot currently be reduced to a credible winner. In randomized obesity studies that directly reported a whole-body breakdown, approximately 25% of weight lost with pooled tirzepatide doses and 26.9% with pooled orforglipron dose groups was lean mass. STEP 1 supports two reproducible semaglutide calculations—38.6% under the treatment-policy estimand and 40.0% under the trial-product estimand—but those are shares of measured fat-plus-lean tissue loss, not shares of scale weight loss and not percentages of skeletal muscle.
The cleanest quotable finding: Among currently approved obesity medicines, randomized studies directly reported a whole-body lean share of approximately 25% for pooled tirzepatide doses and 26.9% for pooled orforglipron dose groups. No randomized head-to-head body-composition trial of the approved products has established which preserves the most lean tissue.
| Ingredient and product context | Primary obesity evidence | Body-composition sample | Dose and duration | Method | Lean result | Lean share | What the number can support |
|---|---|---|---|---|---|---|---|
| Semaglutide — Wegovy injection context | STEP 1 DXA substudy | 140 total; 95 semaglutide, 45 placebo | 2.4 mg weekly; 68 weeks | Whole-body DXA | Treatment-policy: fat −8.36 kg, lean −5.26 kg. Trial-product: fat −10.40 kg, lean −6.92 kg | 38.6% treatment-policy; 40.0% trial-product |
RXI-CALC Two valid calculations from the same DXA group under two different estimands. The denominator is measured fat plus lean tissue loss—not scale weight loss. |
| Tirzepatide — Zepbound context | SURMOUNT-1 DXA substudy | 160 paired scans; 124 tirzepatide, 36 placebo | Pooled 5, 10 and 15 mg; 72 weeks | Whole-body DXA | Body weight −21.3%; fat mass −33.9%; lean mass −10.9% | ~25% | DR-WB Direct whole-body result, pooled across three doses rather than reported separately by dose. |
| Orforglipron — Foundayo context | ATTAIN-1 DXA subgroup | 171 total in the body-composition subgroup | Publication groups pooled; 72 weeks | DXA | Total fat mass −13.8%; lean mass −4.5%; visceral fat −19.0% | 26.9% | DR-WB Direct whole-body result, pooled across publication-dose groups. Current marketed dose numbers are different and are not treated as interchangeable. |
| Liraglutide — Saxenda context | Randomized 20-week obesity study | 564 in the parent trial; exact 3.0 mg body-composition count not stated in the published abstract | 3.0 mg daily; 20 weeks | DXA/CT body-composition substudy | Body fat −15.4%; lean tissue −2.0% | NR | The source reports separate fat and lean changes, not an equivalent dose-specific share that belongs beside the direct tirzepatide and orforglipron values. |
Source line: The RX Index GLP-1 Lean-Mass Evidence Map v1.0, verified August 1, 2026. Primary sources: STEP 1 supplementary appendix, SURMOUNT-1 DXA substudy, ATTAIN-1, and the liraglutide randomized study.
What The RX Index found by assembling the sources
The core approved-drug evidence map contains four active ingredients. It has:
- Two directly reported whole-body shares: tirzepatide and orforglipron.
- One ingredient with two reproducible same-population calculations under different estimands: semaglutide in STEP 1.
- One ingredient with a dose-specific fat-and-lean result but no equivalent share: liraglutide 3.0 mg.
- No randomized head-to-head body-composition result between two currently approved obesity products that establishes a winner.
That is the state of the evidence as of August 1, 2026. The data are real. The clean drug hierarchy people want from them is not.
What does the GLP-1 lean-mass data show—and what does it not show?
The studies show that fat mass generally fell more than lean mass, so lean tissue became a larger proportion of a smaller body even when its absolute amount declined. They do not establish that one approved drug is better for skeletal muscle, that every kilogram of measured lean loss was muscle, that strength declined, or that a trial average predicts an individual result.
What it shows:
- Lean mass declined in several randomized body-composition analyses while fat mass declined more.
- The directly reported whole-body shares for pooled tirzepatide and pooled orforglipron were close: approximately 25% and 26.9%.
- STEP 1 produces different semaglutide shares when the trial's two prespecified estimands are used.
- Evidence coverage is uneven across molecules, formulations and dose levels.
- Whole-body DXA, regional MRI and direct muscle imaging answer different questions.
What it does not show:
- A best-to-worst drug order.
- The percentage of skeletal muscle a person lost.
- A diagnosis of sarcopenia.
- A causal link between a lean-mass change and weakness, falls or disability.
- A personal forecast.
- A reason to start, stop or switch a prescription without a prescriber.
No amount of table formatting makes unlike trials statistically comparable. The point of this page is to show the numbers without hiding the seams.
Why does STEP 1 produce both 38.6% and 40.0% for semaglutide?
Both STEP 1 calculations are reproducible because the trial reported two different estimands. Under the treatment-policy estimand, semaglutide participants lost an estimated 8.36 kg of fat and 5.26 kg of lean body mass; under the trial-product estimand, the estimates were 10.40 kg and 6.92 kg. The answer changes because the analysis question changes—not because the arithmetic is mysterious.
| STEP 1 analysis | What the estimand asks | Fat-mass loss | Lean-body-mass loss | The RX Index calculation | Lean share of measured fat-plus-lean loss |
|---|---|---|---|---|---|
| Treatment-policy estimand | Effect regardless of treatment discontinuation or rescue intervention | 8.36 kg | 5.26 kg | 5.26 ÷ (8.36 + 5.26) | 38.6% |
| Trial-product estimand | Effect if participants remained on treatment as intended without rescue intervention | 10.40 kg | 6.92 kg | 6.92 ÷ (10.40 + 6.92) | 40.0% |
Source line: STEP 1 supplementary appendix, trial-product DXA results on page 20 and treatment-policy DXA results in Table S5 on page 24. Both analyses use 95 semaglutide and 45 placebo participants.
The formula is simple:
The denominator needs a name because it is easy to overstate. It is measured fat-plus-lean tissue loss. It is not total scale weight loss. DXA's compartments do not account cleanly for every kilogram of scale change, and the modeled body-composition values should not be silently paired with a headline weight value from a different analysis population.
The two shortcuts that fail
Do not divide −9.7% lean-mass change by −15.0% body-weight change. The percentages use different baselines: baseline lean mass for one and baseline body weight for the other.
Do not divide a DXA-subgroup lean-mass value by the full-trial weight-loss value. STEP 1's body-composition estimates come from 95 semaglutide participants in the DXA subgroup. The trial's headline weight result comes from a much larger analysis population. A fraction built from those two populations looks exact and means nothing.
This is the section we would want if we were the writer instead of the publisher: the "about 40%" statement can be reproduced, but only after the denominator and estimand are named. It still is not "40% muscle."
How was the evidence map built?
The RX Index reviewed primary trial publications, supplementary appendices, FDA prescribing information, trial records and qualifying congress materials, then extracted each result with its population, dose, duration, body region, estimand and source status intact. We did not collect patient-level data, pool the trials statistically, or convert an incompatible result merely to fill a table cell.
Scope of the core table
The core table includes randomized adult obesity evidence for the current U.S. chronic-weight-management ingredients with GLP-1 activity:
- Semaglutide
- Tirzepatide, which is a dual GIP/GLP-1 receptor agonist
- Liraglutide
- Orforglipron
A result entered the core table only when the source allowed the treatment group, population, duration, method and body region to be identified.
Evidence kept in separate tiers
The following are useful, but they do not belong on the same line as the core approved-drug evidence:
- Investigational medicines or combinations
- Type 2 diabetes populations used outside an obesity-treatment evidence question
- Regional MRI rather than whole-body measurement
- Conference-only analyses
- Prospective single-arm studies
- Observational comparisons and preprints
- Direct muscle imaging rather than DXA lean mass
Those results appear later, labeled for what they are.
Provenance codes
| Code | Meaning |
|---|---|
| DR-WB | The investigators directly reported a whole-body share of measured weight loss attributable to lean mass or lean soft tissue. |
| DR-R | The investigators directly reported, or directly supplied the inputs for, a regional share. It is not a total-body result. |
| RXI-CALC | The RX Index calculated the value from compatible primary-source inputs in the same analyzed group, estimand, timepoint and body region. |
| NR | The source did not report an equivalent result. |
| NQL | No qualifying result was located under this protocol as of the verification date. |
The five-part denominator rule
A calculated share is publishable only when the numerator and denominator come from the same:
- Analyzed participants
- Estimand or missing-data framework
- Timepoint
- Treatment group
- Body region
Fail any one of those five and the honest answer is not reported, not a back-calculation.
Source hierarchy
- Peer-reviewed primary trial publication and supplement
- Current FDA prescribing information for product and dose status
- Trial registry record or results
- Primary congress presentation when a full body-composition paper is unavailable
- Preprint or observational report in a clearly separate evidence tier
Manufacturer medical-information pages and secondary reviews were used as discovery tools, not as substitutes when the primary trial or regulatory source was available.
Negative evidence searches
"No qualifying result located" is a dated search finding, not a claim that no study can exist. Each NQL entry records the medicine, formulation, outcome sought, source set reviewed and verification date. The wording stays narrow on purpose.
Two-pass verification
Every public number was checked once during extraction and again during the final audit. The second pass checked the sign, decimal, unit, sample, treatment group, timepoint, body region, estimand, direct-versus-calculated status and any published correction.
That last step matters. The SURMOUNT-1 body-composition paper has a published correction stating that the Figure 4 y-axis should read kilograms rather than percent. The table values used here are not changed by that correction, but the correction belongs in the source record.
What does the closest same-protocol active comparison show?
REDEFINE 1 provides the closest same-protocol active comparison involving a GLP-1 medicine located in this review. In a 252-person DXA analysis, CagriSema, semaglutide and cagrilintide were measured within the same randomized parent trial; the active-arm lean-soft-tissue shares were 33.1%, 30.3% and 37.1%. The result is more informative than a cross-trial juxtaposition, but it still does not establish a winner because the body-composition analysis was small, congress-reported and did not show pairwise inferential tests of the lean shares.
| Randomized arm | Scale-weight change in DXA subgroup | Fat share of weight loss | Lean-soft-tissue share | Status on August 1, 2026 |
|---|---|---|---|---|
| CagriSema 2.4 mg/2.4 mg | −23.9% | 66.9% | 33.1% | Investigational combination |
| Semaglutide 2.4 mg | −16.6% | 69.7% | 30.3% | Approved ingredient and dose context |
| Cagrilintide 2.4 mg | −15.0% | 62.9% | 37.1% | Investigational |
| Placebo | −2.8% | 56.7% | 43.3% | Control |
Source line: Ravussin E et al., REDEFINE 1 body-composition analysis presented at ECO 2026. The parent randomized trial was published in the New England Journal of Medicine. Values use the trial-product estimand.
The descriptive span across the three active arms is 6.8 percentage points. That is a real observation from one protocol. It is not proof that cagrilintide causes more lean loss than semaglutide, because the presentation does not report that pairwise statistical test and the analysis was not designed as a definitive lean-preservation comparison.
REDEFINE 1 also corrects a common overstatement in the other direction: functional evidence is sparse, but it is not zero. The 30-second sit-to-stand score changed by 2.5 repetitions with CagriSema and 2.4 with placebo; the estimated difference was 0.03, with a 95% confidence interval from −0.7 to 0.8, and was not significant. Patient-reported physical-function measures improved more with CagriSema.
Which supporting studies should not be mixed into the approved-drug table?
Several useful results answer a different question from the core table. Regional MRI, type 2 diabetes substudies, an investigational triple agonist and a conference-only Taiwanese subgroup all add evidence, but mixing them into one numerical range would create exactly the false precision this page is designed to prevent.
| Treatment and study | Population | Sample and duration | Method or scope | Verified result | Why it stays separate |
|---|---|---|---|---|---|
| Semaglutide 2.4/7.2 mg pooled — STEP UP | Obesity without diabetes | 55 total; 49 semaglutide, 6 placebo; 72 weeks | MRI from thoracic vertebra 9 to the knees; volume converted to mass | 15.6% of measured regional mass reduction was lean tissue; 84.4% was adipose tissue | DR-R Regional, not whole-body; doses pooled; very small placebo group; congress analysis. |
| Semaglutide 1.0 mg — SUSTAIN 8 | Type 2 diabetes on metformin | 53 semaglutide participants with observed end-of-treatment data; 52 weeks | Whole-body DXA | Fat −3.4 kg; lean −2.3 kg; 40.4% RXI-CALC | Diabetes dose and population; active non-GLP-1 comparator; no placebo arm. |
| Canagliflozin 300 mg — SUSTAIN 8 control | Type 2 diabetes on metformin | 61 with observed end-of-treatment data; 52 weeks | Whole-body DXA | Fat −2.6 kg; lean −1.5 kg; 36.6% RXI-CALC | Non-GLP-1 active control. Between-group body-composition changes were not significantly different. |
| Retatrutide 4/8/12 mg pooled — phase 2 substudy | Type 2 diabetes | 103 end-of-treatment scans; 36 weeks | Whole-body DXA | Reported fat-loss index 64.6%; complement 35.4% | Investigational, pooled doses, short duration and diabetes population. |
| Semaglutide 2.4 mg — STEP 12 | Chinese adults with overweight or obesity | 45-person Taiwan body-composition subset; 44 weeks | Body-composition analysis | Total-fat-mass proportion −6.2 percentage points; lean-body-mass proportion +5.7 points | The accessible primary abstract does not report compatible absolute fat and lean losses, so no share is calculated. |
Source line: STEP UP congress analysis, SUSTAIN 8, retatrutide body-composition substudy, and STEP 12 congress abstract.
Why the STEP UP calculation is legitimate but not interchangeable
The STEP UP scan covered the region from T9 to the knees. It excluded the head, arms, neck and lower legs. The investigators converted tissue volumes to mass using densities of 0.92 for adipose tissue and 1.1 for lean tissue, then reported the 84.4%/15.6% split.
That is a carefully documented result. It is also not a total-body result. It belongs in the evidence map, not on the same visual axis as whole-body DXA.
Why the SUSTAIN 8 sample needed correcting
The substudy enrolled 178 people—88 assigned to semaglutide and 90 to canagliflozin. The confirmatory observed end-of-treatment analysis included 114 total, split into 53 semaglutide and 61 canagliflozin participants. "114" is not the sample size for each arm.
Which current products and doses have body-composition evidence?
Current U.S. product coverage is patchier than the amount of commentary suggests. Zepbound has a pooled whole-body DXA share across its three maintenance doses; Wegovy injection has STEP 1 at 2.4 mg; Saxenda has a 3.0 mg fat-and-lean result without an equivalent share; Foundayo's publication-dose labels do not numerically match its marketed strengths; and no qualifying whole-body result was located for Wegovy tablets or a dose-specific Wegovy HD 7.2 mg group.
| Product or formulation | Current label context on August 1, 2026 | Closest body-composition evidence | Coverage status |
|---|---|---|---|
| Saxenda — liraglutide injection | 3.0 mg once daily | Randomized 20-week liraglutide 3.0 mg result | Fat and lean changes reported; equivalent dose-specific lean share NR |
| Wegovy injection — semaglutide | 1.7 or 2.4 mg weekly maintenance under the standard injection label | STEP 1 at 2.4 mg | Whole-body STEP 1 values support 38.6% and 40.0% RXI calculations under separate estimands |
| Wegovy HD — semaglutide injection | FDA-approved 7.2 mg higher-dose product on March 19, 2026 | STEP UP pooled 2.4/7.2 mg regional MRI | Dose-specific 7.2 mg whole-body share NQL under this protocol |
| Wegovy tablets — semaglutide | 25 mg once-daily maintenance after 1.5, 4 and 9 mg escalation steps | OASIS obesity evidence reviewed | Qualifying quantitative whole-body lean-share result NQL under this protocol |
| Zepbound — tirzepatide injection | 5, 10 or 15 mg once-weekly maintenance | SURMOUNT-1 pooled 5/10/15 mg DXA | Direct whole-body share available; not dose-specific |
| Foundayo — orforglipron tablet | 0.8 mg start, then 2.5 and 5.5 mg; optional steps to 9, 14.5 and 17.2 mg; maximum 17.2 mg | ATTAIN-1 pooled publication groups reported as 6, 12 and 36 mg | Direct whole-body share available, but publication-dose and marketed-dose numbers are not assumed interchangeable |
Source line: Current product and dose status from FDA prescribing information and FDA announcements: Wegovy tablets and standard injection, Wegovy HD approval, Zepbound prescribing information, Foundayo prescribing information, and Saxenda prescribing information.
Searchers use "GLP-1" as an umbrella term. Pharmacologically, tirzepatide is a dual GIP/GLP-1 receptor agonist. That distinction stays visible here because the molecule's mechanism should not be blurred merely to match everyday language.
What do the 2026 meta-analyses add?
A 2026 synthesis of 20 randomized trials found that lean mass represented roughly one-quarter to just over one-third of total weight lost with the three incretin therapies it analyzed. The pooled figures were 35.2% for semaglutide, 25.4% for tirzepatide and 26.8% for liraglutide, but these remain indirect cross-trial estimates rather than a randomized medicine-versus-medicine comparison.
| Weight-loss approach | Pooled lean share of total weight lost | 95% confidence interval |
|---|---|---|
| Semaglutide | 35.2% | 31.5%–38.9% |
| Tirzepatide | 25.4% | 22.8%–28.0% |
| Liraglutide | 26.8% | 23.1%–30.5% |
| Lifestyle intervention | 26.2% | 24.1%–28.3% |
| Lifestyle plus resistance training | 17.5% | 14.2%–20.8% |
Source line: Eisa N, Barood O. Lean Mass Changes With Incretin Therapy Versus Lifestyle Intervention. Twenty randomized trials, 15,782 participants; heterogeneity I² = 68%.
This is useful context, not permission to turn the three pooled values into a causal drug order. Each medicine's estimate is assembled from a different trial mix.
A separate 2026 International Journal of Obesity meta-analysis of seven studies and 821 patients found two results that sound contradictory until their denominators are separated:
- Lean mass as a proportion of body weight increased by 1.81 percentage points (95% CI 1.10–2.52; I² = 7%).
- Absolute lean mass decreased by 1.74 kg (95% CI −3.04 to −0.45; I² = 98%).
The pooled results show that fat fell faster than lean tissue overall, improving relative body composition. Absolute lean-mass estimates were highly heterogeneous. That is the statistical version of what the evidence tables show row by row.
Is lean mass the same as muscle?
No. DXA lean soft tissue is a body-composition compartment, while skeletal muscle is an anatomically defined tissue. A DXA-measured lean decline may include changes in body water, protein, glycogen, organs and other soft tissues; it cannot be relabeled one-for-one as lost skeletal muscle.
| What the source measured | Accurate wording | Do not write |
|---|---|---|
| DXA lean mass or lean soft tissue | "DXA-measured lean mass declined by X" | "The person lost X of skeletal muscle" |
| Regional MRI lean tissue | "Lean tissue declined in the scanned region" | "X% of total-body weight loss was muscle" |
| MRI thigh muscle volume | "Thigh muscle volume decreased" | "Total lean body mass decreased" |
| Handgrip or sit-to-stand | "Measured function improved, declined or did not differ" | "The imaging result proves strength changed" |
| Fat-free mass | "Fat-free mass changed by X" | Treating fat-free mass, lean soft tissue and skeletal muscle as identical |
Source line: Heymsfield SB et al., Body Composition: A Critical Component of the Definition of Sarcopenia, which separates molecular fat-free or lean compartments from organ- and tissue-level skeletal muscle.
This distinction is not semantic housekeeping. A headline can turn "DXA lean mass fell" into "patients lost muscle" in one sentence, and the second statement is more specific than the measurement allows.
Does lean-mass loss mean strength or physical function declined?
Not automatically. Functional evidence is sparse, but it is not zero. REDEFINE 1 found no significant CagriSema-versus-placebo difference in sit-to-stand change, while the prospective single-arm SEMALEAN study reported a 4.5 kg improvement in handgrip strength at 12 months despite an earlier lean-mass decline.
| Study | Design and population | Tissue result | Function or muscle-quality result | What it can support |
|---|---|---|---|---|
| REDEFINE 1 | Randomized obesity trial; CagriSema versus active arms and placebo | Lean soft tissue declined while its body-weight proportion increased | Sit-to-stand change 2.5 with CagriSema vs 2.4 with placebo; ETD 0.03 (95% CI −0.7 to 0.8), not significant. Patient-reported physical function improved more with CagriSema | Objective sit-to-stand did not worsen relative to placebo in this analysis; patient-reported function improved |
| SEMALEAN | Prospective single-arm semaglutide 2.4 mg study; 115 enrolled, 106 completed | Lean mass −3 kg at month 7, then stabilized | Handgrip strength +4.5 kg at month 12; sarcopenic-obesity prevalence 49% to 33% | Function improved during follow-up, but the design cannot isolate a drug effect without a control group |
| SURPASS-3 MRI | Post hoc randomized active-comparator analysis in type 2 diabetes | Thigh muscle volume declined | Muscle fat infiltration improved; muscle-volume decline was broadly consistent with the amount of weight lost | Direct regional muscle imaging adds quality information, but it is not a whole-body obesity-trial lean-share result |
Source line: REDEFINE 1 ECO 2026 analysis, SEMALEAN, and SURPASS-3 MRI.
The evidence gap is not "nobody has measured function." The gap is that objective functional outcomes are uncommon, often secondary, and especially thin in older or frail populations where losing tissue may matter most.
For an individual patient, the relevant question is not merely what happened to a trial mean. Baseline muscle reserve, age, frailty, nutrition, activity, disease burden and the amount and speed of weight loss all matter. That is a prescriber-level assessment, not something a percentage on this page can perform.
Can body composition be changed without giving up weight loss?
The BELIEVE phase 2 trial shows that the fat-versus-lean split is biologically modifiable, at least with an investigational combination. At week 72, semaglutide 2.4 mg reduced body weight by 15.7% and lean mass by 7.4%, while high-dose bimagrumab plus semaglutide reduced body weight by 22.1% and lean mass by 2.9%; 92.2% of the combination's weight loss was attributed to fat mass.
| Treatment | Body-weight change | Fat-mass change | Lean-mass change | Fat share of weight loss |
|---|---|---|---|---|
| Bimagrumab 30 mg/kg | −10.8% | −28.5% | +2.5% | 100% |
| Semaglutide 2.4 mg | −15.7% | −27.8% | −7.4% | 75.6% |
| High-dose bimagrumab + semaglutide | −22.1% | −45.7% | −2.9% | 92.2% |
Source line: Heymsfield SB et al., Bimagrumab plus semaglutide alone or in combination for the treatment of obesity, a randomized phase 2 trial published in Nature Medicine in 2026.
This does not turn bimagrumab into a recommendation. It is an investigational intravenous agent, this was a phase 2 program, and the week-72 figures use an efficacy estimand. What the trial establishes is narrower and still important: the composition of drug-associated weight loss is not a fixed law of nature.
Why do published GLP-1 lean-mass percentages disagree?
The numbers disagree because studies report different outcomes, use different denominators and measure different bodies in different ways. A percentage can describe change from baseline, share of tissue lost, share of scale weight lost, or change in the proportion of the body that is lean. Those are four different quantities.
Different denominators
- Percent change in lean mass: change relative to starting lean mass.
- Lean share of measured fat-plus-lean loss: lean loss divided by fat loss plus lean loss.
- Lean share of scale-weight loss: lean loss divided by total scale-weight loss.
- Percentage-point change in lean proportion: change in lean mass as a percentage of current body mass.
The same participant can lose absolute lean mass while the proportion of the body that is lean rises. Fat only has to fall faster.
Different measurement methods
DXA estimates fat mass, bone mineral and lean soft tissue. MRI can quantify regional adipose tissue, organ volumes or specific muscle groups depending on the protocol. A scan from T9 to the knees is not a whole-body scan. Bioelectrical impedance introduces another set of assumptions.
Different populations
Obesity without diabetes, type 2 diabetes, older age, low baseline muscle reserve and prior weight loss are not interchangeable populations. SUSTAIN 8 and the retatrutide substudy involved type 2 diabetes; STEP 1, SURMOUNT-1 and ATTAIN-1 were obesity studies without diabetes in the relevant body-composition analyses.
Different treatment exposure
Trial-product, efficacy, treatment-policy and treatment-regimen estimands answer different questions about adherence, rescue therapy and missing data. STEP 1 demonstrates that the estimand alone can move the calculated share from 38.6% to 40.0% without changing the molecule or dose.
Pooled doses
SURMOUNT-1 pools tirzepatide 5, 10 and 15 mg. ATTAIN-1 pools its publication-dose groups for body composition. STEP UP pools semaglutide 2.4 and 7.2 mg. A pooled figure should not be assigned to one maintenance dose.
Which drug causes the most or least lean-mass loss?
No approved product can be named from the current randomized body-composition evidence. The direct whole-body values for pooled tirzepatide and pooled orforglipron are numerically close, while semaglutide's STEP 1 calculation changes with the estimand and liraglutide lacks an equivalent dose-specific share. A numerical order built across those sources would imply a controlled comparison that did not occur.
The closest same-protocol GLP-1-based comparison is REDEFINE 1, where active-arm lean-soft-tissue shares ranged from 30.3% to 37.1%. That analysis includes investigational cagrilintide and CagriSema, has 252 participants across the DXA subgroup, and does not show pairwise inferential testing of the lean shares.
A 2026 routine-care preprint compared semaglutide and tirzepatide observationally and associated tirzepatide with 1.1 to 2.0 percentage points greater relative lean-body-mass decline across assessed timepoints. The paired body-composition analysis included 7,965 patients, but treatment was not randomized, clinical measurements were heterogeneous, and the report was not peer reviewed at the verification date. It generates a comparative signal. It does not establish that tirzepatide caused more loss.
A clean answer would require a randomized active comparison of approved products with the same whole-body protocol, dose-specific reporting, prespecified function measures and enough participants to detect a modest between-treatment difference. That study was not located.
What evidence is still missing?
The biggest gaps are not hidden in a footnote: dose-specific whole-body results, approved-product head-to-head evidence, objective function in older and frail adults, and current-formulation coverage are still missing.
| Evidence gap | Why it matters | Refresh trigger |
|---|---|---|
| No randomized head-to-head whole-body comparison between two approved obesity products located | Prevents a controlled answer to which approved medicine preserves the most lean tissue | Publication of an active-comparator DXA/MRI trial with aligned methods |
| Tirzepatide body composition pooled across 5, 10 and 15 mg | The approximately 25% result cannot be assigned to one dose | Dose-specific SURMOUNT body-composition publication |
| Orforglipron body composition pooled across publication-dose groups | The 26.9% result cannot be assigned to a current marketed strength | Dose-specific ATTAIN publication or FDA review detail |
| No qualifying whole-body result located for Wegovy tablets 25 mg | Injectable evidence should not be silently transferred to an oral formulation | OASIS body-composition analysis or label supplement |
| Wegovy HD evidence pooled with 2.4 mg and measured regionally | The 15.6% result is not a dose-specific whole-body 7.2 mg estimate | Dose-specific total-body DXA/MRI publication |
| Sparse objective function in older, frail and low-muscle populations | A tissue change does not by itself establish weakness or disability | Trials reporting strength, gait, sit-to-stand and clinical function in those populations |
| REDEFINE 1 body-composition detail is congress-reported | Full peer review may add analyses, corrections or pairwise tests | Peer-reviewed body-composition paper or regulatory review |
| Observational semaglutide-versus-tirzepatide comparison remains a preprint | Conclusions may change after peer review or revision | Journal publication, correction or revised preprint |
Source line: The RX Index evidence-gap log v1.0. "No qualifying result located" means no result meeting this page's stated protocol was found by August 1, 2026.
Limitations
We would rather you know these than discover them.
- The core comparisons are indirect. Putting values in adjacent rows does not make them randomized against one another.
- The body-composition substudies are small. STEP 1's DXA analysis includes 95 semaglutide and 45 placebo participants. SURMOUNT-1's paired DXA analysis includes 160 total. ATTAIN-1's DXA subgroup includes 171. Those are useful samples, but far smaller than the parent efficacy trials.
- Some results are pooled across doses. Tirzepatide, orforglipron, retatrutide and STEP UP do not provide the displayed composition share separately for each dose.
- Some results come from a conference presentation. REDEFINE 1 body composition, STEP UP and STEP 12 may be expanded or revised in full publication.
- DXA lean mass is sensitive to hydration and is not a direct muscle measurement. Rapid weight loss can change body water, and lean soft tissue includes more than skeletal muscle.
- Regional MRI and whole-body DXA are not interchangeable. STEP UP's T9-to-knee result is kept outside the core table for that reason.
- The same word can hide different estimands. Treatment-policy and trial-product estimates answer different questions and should not be mixed.
- Evidence in older and frail adults is limited. Trial means should not be treated as guarantees for people with low baseline muscle reserve, heart failure, chronic kidney disease or established frailty.
- The RX Index calculations are calculations. The equations and primary inputs are shown so they can be checked. If an input or arithmetic step is wrong, the error is ours.
- This page is a structured evidence map, not a patient-level meta-analysis or a medical recommendation. It cannot tell an individual how much lean tissue they will lose or whether a medication is appropriate.
Frequently asked questions
How to cite this page
Suggested citation
Figures labeled DR-WB or DR-R are reported by the original investigators. Figures labeled RXI-CALC are calculations by The RX Index using the formula and primary-source inputs shown on this page.
Primary sources
Core approved-drug evidence
- Wilding JPH, Batterham RL, Calanna S, et al. Once-Weekly Semaglutide in Adults with Overweight or Obesity. New England Journal of Medicine. 2021;384:989–1002. Body-composition estimands are reported in the supplementary appendix.
- Look M, Dunn JP, Kushner RF, et al. Body composition changes during weight reduction with tirzepatide in SURMOUNT-1. Diabetes, Obesity and Metabolism. 2025;27:2720–2729. Correction: doi:10.1111/dom.70050.
- Wharton S, et al. Orforglipron for obesity treatment in ATTAIN-1. New England Journal of Medicine. 2025.
- Astrup A, Carraro R, Finer N, et al. Safety, tolerability and sustained weight loss over 2 years with liraglutide. International Journal of Obesity. 2012;36:843–854.
Same-protocol, regional and supporting trial evidence
- Garvey WT, et al. Cagrilintide–semaglutide for obesity in REDEFINE 1. New England Journal of Medicine. 2025;393:635–647. Body-composition analysis: ECO 2026 presentation.
- Hjelmesaeth J, et al. STEP UP body-composition and function analysis. EASD 2025 congress presentation.
- McCrimmon RJ, Catarig AM, Frias JP, et al. Semaglutide versus canagliflozin body composition in SUSTAIN 8. Diabetologia. 2020;63:473–485.
- Effects of retatrutide on body composition in people with type 2 diabetes. The Lancet Diabetes & Endocrinology. 2025.
- Guo L, et al. STEP 12 body-composition analysis. Obesity. 2025 congress abstracts.
- Sattar N, Neeland IJ, Dahlqvist Leinhard O, et al. Tirzepatide and muscle composition changes in SURPASS-3 MRI. The Lancet Diabetes & Endocrinology. 2025;13:482–493.
Function, synthesis and body-composition interpretation
- Alissou M, Demangeat T, Folope V, et al. The SEMALEAN study. Diabetes, Obesity and Metabolism. 2026;28:112–121.
- Eisa N, Barood O. Lean Mass Changes With Incretin Therapy Versus Lifestyle Intervention. Diabetes, Obesity and Metabolism. 2026;28:4818–4827.
- Laverde LP, et al. Effect of GLP-1 receptor agonists at obesity-management doses on muscle health. International Journal of Obesity. 2026.
- Heymsfield SB, et al. Bimagrumab plus semaglutide alone or in combination for obesity. Nature Medicine. 2026.
- Heymsfield SB, et al. Body Composition: A Critical Component of the Definition of Sarcopenia. Journal of the Endocrine Society. 2024;8:bvae164.
- Murugadoss K, Venkatakrishnan A, Soundararajan V. Greater lean-body-mass decline with tirzepatide than semaglutide in routine care. medRxiv. 2026. Preprint; not peer reviewed at the verification date.
Current U.S. regulatory sources
- FDA. Wegovy injection and tablets prescribing information. Revised 2026.
- FDA. Approval of Wegovy HD 7.2 mg. March 19, 2026.
- Eli Lilly. Zepbound prescribing information. Current at verification.
- FDA. Foundayo prescribing information. Revised April 2026.
- FDA. Foundayo approval announcement. April 1, 2026.
- FDA. Saxenda prescribing information. Current at verification.
Related research
Change log
Version history
August 1, 2026 — Version 1.0. Initial publication. The core approved-drug map covers four active ingredients: two directly reported whole-body shares, two STEP 1 same-population calculations under separate estimands, and one dose-specific liraglutide result without an equivalent share. Separate tiers cover same-protocol investigational evidence, regional imaging, type 2 diabetes substudies, functional outcomes, observational comparisons and meta-analytic context.