Key Takeaways
- A real-world analysis published 12 November 2026 in npj Exercise Medicine and Health compared lean-mass change during tirzepatide versus semaglutide use in routine care, screening 670,422 first-time users and identifying 7,965 with paired measurements across 12 months.
- In raw, unadjusted comparisons, tirzepatide was associated with greater relative lean-body-mass loss than semaglutide over 12 months, with a between-agent difference of −2.1 percentage points (95% CI −3.8 to −0.3).
- After statistical adjustment for concurrent total-weight change, that difference disappeared, shrinking to −0.1 percentage points (95% CI −1.6 to 1.4) — an estimate indistinguishable from zero, suggesting the raw gap tracked weight-loss depth rather than agent identity.
- Lean body mass accounted for approximately 20% of the total weight lost with either agent, meaning roughly four-fifths of recorded loss was non-lean tissue rather than muscle-containing compartments.
- An exploratory digital phenotype called Depletive — more than 20% body-weight loss alongside more than 5% lean-mass loss — was more frequent with tirzepatide (10.3% versus 6.7%), while the favorable Prime phenotype appeared at nearly identical rates (11.8% versus 12.3%).
- Because the analysis is observational and used routine electronic health records rather than randomized assignment, its findings describe statistical associations among compound choice, weight-loss depth, and lean-mass trajectories; they cannot establish causation.
The direct answer: In a large real-world analysis published 12 November 2026 in npj Exercise Medicine and Health, tirzepatide was associated with greater relative lean-body-mass loss than semaglutide over 12 months of routine care — a raw difference of −2.1 percentage points. In absolute terms the gap was small, and after adjustment for concurrent total-weight change it vanished entirely (−0.1 percentage points, 95% CI −1.6 to 1.4). Lean tissue represented roughly 20% of the weight lost on either agent. The study was explicitly observational: it mined electronic health records, not a randomized trial, so the numbers describe associations between weight-loss depth and body-composition change — not causal effects of one compound over another. (For compound-by-compound context, see BioPharma’s three-compound comparator explainer.)
Research boundary: This article is written for research education only. BioPharma laboratory reference materials are manufactured and sold for qualified in-vitro research use and are not intended for human or veterinary use of any kind. No dosing, preparation, route, administration, treatment, efficacy, or safety instructions appear here, and none should be inferred. The figures that follow describe published clinical-research records — they refer to analyzed cohorts in the literature, not to catalog products.
What the 2026 Real-World Analysis Did
The study — 10.1038/s44437-026-00025-y, published open-access in npj Exercise Medicine and Health on 12 November 2026 — applied an EHR-linked, LLM-enabled body-composition phenotyping pipeline to routine electronic health records. Starting from 670,422 first-time users of semaglutide (456,742) or tirzepatide (213,680), the pipeline scanned structured data and free-text clinical documentation for body-composition measurements falling in defined windows around treatment initiation. The paired analytic subset contained 7,965 patients with usable readings both before and within 12 months after starting therapy; within that group, 3,050 semaglutide records and 858 tirzepatide records contributed landmark lean-mass values. Paired measurements matter because each patient acts as their own baseline: percent change can be computed within a person rather than across unlike cohorts, which narrows some confounding even in non-randomized data.
That subset represents about 1.2% of first-episode users, and the selection caveat is worth naming: patients who happened to carry repeated body-composition data likely received more intensive monitoring than typical recipients, so heavier clinical attention may be overrepresented at both ends of the outcome spectrum. Measurement modality leaned heavily on bioelectrical-impedance scales — Tanita devices supplied 67% of the readings — while DEXA scans and Bod Pod air-displacement tests each contributed under 1%. Modality matters for interpretation, because impedance estimates are sensitive to hydration status. As the study defines it, lean body mass includes body water and non-muscle lean tissue such as organs and connective structures; it is not a direct readout of skeletal muscle mass, muscle quality, strength, or physical performance — a distinction the authors repeat throughout the paper.
The Head-to-Head Lean-Mass Numbers
This is where the 2026 real-world head-to-head earns its keep. Published in npj Exercise Medicine and Health (10.1038/s44437-026-00025-y), the study places a GLP-1 receptor agonist (semaglutide) and a dual GIP/GLP-1 receptor agonist (tirzepatide) under the same observational microscope and asks which one sheds more lean mass — and whether that answer survives once total weight change is priced in. For readers who want the broader pharmacology laid out side by side first, BioPharma’s three-compound comparator explainer covers the retatrutide–semaglutide–tirzepatide landscape in depth.
The raw difference
The unadjusted reading favors semaglutide. The study headlines lean-mass change as a percentage of each participant’s baseline rather than as absolute kilograms, and at 12 months that relative change was −3.6% (s.d. 10.4) on semaglutide versus −5.6% (s.d. 9.8) on tirzepatide — a between-drug difference of −2.1 percentage points (95% CI −3.8 to −0.3; P = 0.019). On its face, that is a clear edge for the GLP-1 agent. But the same cohort records the trade-off that makes the number hard to read alone: tirzepatide patients lost substantially more weight overall, −15.2% versus −9.7% at 12 months (between-drug −5.5 pp, 95% CI −6.4 to −4.5). Much larger total losses mechanically tilt a proportional lean-mass figure, so the raw comparison is not composition-matched before any biology is even considered.
After the weight-change adjustment
The paper’s key reframing is a two-step adjustment that takes the deck apart. Step one controls the comparison for age, sex, baseline BMI, weight and lean mass, diabetes status, prior GLP-1 exposure, index year and modality; the 12-month between-drug difference eases only slightly, to −1.9 pp (95% CI −3.6 to −0.3), still nominally favoring semaglutide. Step two additionally adjusts for concurrent percent weight change — and the signal disappears outright: −0.1 pp (95% CI −1.6 to 1.4; P = 0.93). Read that second row carefully: whatever separated the two agents on lean mass is fully absorbed by how much weight each group lost. In this cohort, lean-mass loss behaves as a scaled companion of total weight loss rather than a class-specific effect, which is precisely the reframing the paper advances. The same table compresses the story into phenotype shorthand — the “depletive” pattern (>20% weight loss with >5% lean-mass loss) was more frequent on tirzepatide (10.3% vs 6.7%; P < 0.001), while “prime” rates (>10% weight loss, <5% lean-mass loss) were statistically level (11.8% vs 12.3%; P = 0.66).
| Measure | Semaglutide | Tirzepatide | Between-drug difference |
|---|---|---|---|
| Body-weight change at 12 months | −9.7% | −15.2% | −5.5 pp (95% CI −6.4 to −4.5) |
| Relative lean-mass change at 12 months (raw) | −3.6% | −5.6% | −2.1 pp (95% CI −3.8 to −0.3) |
| Adjusted for demographics/baseline covariates | — | — | −1.9 pp (95% CI −3.6 to −0.3) |
| Adjusted for concurrent weight change | — | — | −0.1 pp (95% CI −1.6 to 1.4) |
| Share of weight lost that was lean tissue | ~20% | ~20% | similar between agents |
| Depletive phenotype (>20% weight loss, >5% lean-mass loss) | 6.7% (n = 252) | 10.3% (n = 142) | P < 0.001 |
| Prime phenotype (>10% weight loss, <5% lean-mass loss) | 12.3% (n = 463) | 11.8% (n = 162) | P = 0.66 |
Every figure in Table 1 is transcribed from the published record; pp = percentage points.
What Lean-Mass Share of Weight Loss Means
Strip the head-to-head down to its simplest ratio and one number emerges: on either agent, lean tissue made up roughly one fifth — about 20% — of the weight lost. To judge whether that is reassuring or alarming, the primary authors’ own discussion supplies the calibration: across the published GLP-1 receptor-agonist literature, the lean share of total weight loss spans roughly 15% to 60%, with SURMOUNT-1 (tirzepatide) sitting around one quarter and STEP 1 (semaglutide) around two fifths. Against that band, the real-world ~20% lands toward the favorable end for both compounds. Why the share swings so widely between studies remains an open mechanistic question spanning cellular energetics, muscle-protein turnover, and tissue-biology research.
The interventional counterpoint arrives from the T-REx randomized trial (10.1038/s41591-026-04752-5), which measured lean soft tissue directly with DXA rather than inferring it: losses of −4.67 kg with structured resistance training versus −3.57 kg without, a between-group difference of 1.09 kg (95% CI 0.43–1.75), even though weight loss and fat loss were similar across arms. That dissociation is the point — an interventional lever moved the lean compartment while the scale and the fat compartment barely registered a difference, a separate interventional reading that supports the same quality-of-weight-loss framing the cohort data advance. The cohort’s companion analysis of weight-loss dynamics (10.1093/pnasnexus/pgag171) carries the naturalistic thread further, mapping how these composition outcomes unfold along real-world trajectories.
Who Lost Proportionally More Lean Tissue: the Phenotype Analysis
The Depletive pattern
To isolate patients whose weight reduction came disproportionately at the expense of lean tissue, the analysis defined a combined phenotype: more than 20% total-body-weight loss together with more than 5% lean-mass loss. In this cohort, 10.3% of tirzepatide-treated patients (n = 142) met both thresholds versus 6.7% of semaglutide-treated patients (n = 252), a difference of P < 0.001. Baseline BMI was broadly comparable between the groups, so the gap is not explained by heavier patients concentrating on one side of the comparison, and within this cohort the Depletive pattern was largely confined to women and to patients under 65. The frequencies are eye-catching, but they describe a tail of the distribution carved out by the study’s own thresholds — a point the authors themselves underline by calling the design exploratory.
The Prime pattern
The mirror-image phenotype — more than 10% body-weight loss with less than 5% lean-mass loss — told a more symmetrical story: 11.8% of tirzepatide-treated patients (n = 162) versus 12.3% of semaglutide-treated patients (n = 463), with P = 0.66 and no significant difference between the drugs. One asymmetry is worth carrying forward: mean baseline BMI was higher among semaglutide-treated patients in the Prime group, hinting that starting body composition may shape which pattern a patient drifts toward. Above all, the authors’ own label should govern how these results are quoted: the thresholds are exploratory cut-points chosen for analysis, not validated clinical categories, and “Depletive” and “Prime” are analytical labels rather than diagnoses or established phenotypes of either medication.
Function Signals and Their Limits
The documentation layer added a functional dimension to the body-composition signal. Clinicians recorded reduced exercise tolerance 7.2 percentage points more often among semaglutide-treated patients with greater lean-mass loss, and 11.1 percentage points more often among tirzepatide-treated patients with greater lean-mass loss, while mentions of fatigue rose along a similar gradient. Read quickly, those numbers invite the conclusion that patients losing more lean tissue were measurably losing the capacity to move — a conclusion the underlying material cannot carry.
What the analysis actually relates is a body-composition measurement to free-text clinical documentation. No objective strength or performance testing was available anywhere in the records. Of the labels examined, only the exercise-tolerance term was validated, with a positive predictive value of 100% and a miss rate of 10% among term-bearing notes; the other labels were never put through that validation, so the fatigue signal in particular rests on unverified pattern-matching. Manual review complicated the picture further: the documented limitation was most often musculoskeletal pain limiting walking or activity — a complaint about joints and discomfort, not a measured decline in muscle function. The authors explicitly note that reverse causation, in which patients who already moved less lose more lean tissue, and differences in documentation practices are equally compatible with what was observed. The defensible reading is hypothesis-generating correlation, not evidence of functional decline.
Where the Evidence Does Not Reach
- No causal ranking: The observational design precludes causal inference, so the comparison cannot establish that one agent preferentially depletes lean tissue. Treatment selection, baseline health, diet, physical activity, protein intake and broader lifestyle factors are captured incompletely, if at all, in EHR data.
- Lean mass is not muscle: Impedance-based lean estimates include body water and non-muscle organs, so a decline in estimated lean mass does not by itself establish clinically meaningful sarcopenia. No strength or performance testing was performed in this cohort, leaving functional significance open.
- Measurement heterogeneity: Two-thirds of measurements used Tanita impedance devices, one fifth of patients had no documented device, and only 39% had baseline and follow-up from the same device type. Same-device and Tanita-only restrictions gave similar between-drug results, but hydration and calibration effects on impedance cannot be excluded.
- Persisters and monitoring bias: The paired cohort represents 1.2% of first-episode users and skews toward patients who persisted with treatment and received more monitoring. The landmark estimates therefore describe treated persisters under observation rather than newly treated patients overall, a narrowing that should temper extrapolation.
How This Sits Alongside BioPharma’s Comparator Coverage
Head-to-head weigh-offs are ground BioPharma has already covered in depth: BioPharma’s three-compound comparator explainer lays out which agent produced more weight loss across trials, compound by compound. This article answers a different question — what fraction of that loss is lean tissue, and what the adjustment analysis changes about how those head-to-head figures should be read — so the two pages complement rather than collide. Three nuances from the digital-phenotyping cohort are worth carrying over here. After adjustment, men lost 7–10 percentage points less lean mass than women in this cohort, and the Depletive pattern was concentrated in women under 65. Higher dose and cumulative exposure did track with greater lean loss, but in a way entangled with weight depth — people on higher doses simply lost more weight overall — and the authors explicitly caution it should not be read as an independent pharmacologic dose effect.
For laboratory context, BioPharma lists GLP-3 (Retatrutide) 10mg research reference with batch documentation, and the wider catalog is organized under the peptide research category. Nothing in this real-world clinical analysis transfers any claim to catalog items: the digital-phenotyping cohort describes incretin-class prescription compounds used in routine medical care, while catalog materials are laboratory reference items for in-vitro research. Different use contexts, different evidence standards — and no body-composition outcome observed in clinical care should be read onto any catalog page.
Key Terms and Definitions
- Lean body mass
- The total mass of everything in the body that is not fat, including body water, bone mineral, organs, and non-muscle lean tissue. It is not a direct measure of skeletal muscle, although the two are often conflated in casual commentary.
- Body-composition digital phenotyping
- The pipeline used in the primary study: free-text notes and discrete measurements extracted from electronic health records, then processed by a language-model workflow that derives longitudinal estimates of body weight and lean mass change for large routine-care cohorts.
- BIA
- Bioelectrical impedance analysis, the method behind the Tanita-style scales commonly seen in clinics. A weak electrical current estimates body composition, but readings are sensitive to hydration status, food intake, and recent exercise, which limits precision for tracking small lean-mass shifts.
- Adjusted difference
- The headline −0.1 percentage-point figure describing the difference in lean-mass change between the two agents once concurrent weight change was accounted for statistically. It reframes a seemingly large unadjusted gap as near-equal relative composition change at matched weight loss.
- Depletive phenotype
- An exploratory pattern identifying people who lost more than 20 percent of body weight alongside more than 5 percent lean-mass loss. In this cohort it emerged rarely and was concentrated among women under 65; it requires validation before any clinical use.
- Prime phenotype
- An exploratory pattern describing people who lost more than 10 percent of body weight while keeping lean-mass loss under 5 percent. It represents favorable relative composition change during substantial weight reduction, but remains a descriptive label pending external validation.
Research-Use Boundary and Source Record
Research disclaimer: For research and laboratory use only. Not intended for human or veterinary use. This article does not constitute medical advice. BioPharma provides no dosing, preparation, route, administration, treatment, efficacy, or safety instructions.
- “Greater lean-body-mass decline with tirzepatide than semaglutide in routine care, revealed by body-composition digital phenotyping,” npj Exercise Medicine and Health 1:15 (2026), published 12 November 2026 — primary study.
- “Resistance exercise to attenuate lean mass loss during tirzepatide therapy for obesity: a randomized controlled trial,” Nature Medicine (2026), published 9 November 2026 — T-REx RCT.
- “Weight-loss dynamics with tirzepatide versus semaglutide,” PNAS Nexus 5(6) (2026) — companion real-world cohort analysis.
- “Retatrutide vs Semaglutide vs Tirzepatide: What the Research Actually Shows,” BioPharma comparator explainer.
Frequently Asked Questions
Did the 2026 real-world analysis find that tirzepatide causes greater lean-mass loss than semaglutide?
No. The analysis reported an adjusted between-group difference of −0.1 percentage points (95% CI −1.6 to 1.4), an interval that crosses zero. Because the data were observational — electronic health records rather than randomization — the findings cannot establish that either agent causes greater or lesser lean-mass loss.
What does “lean body mass” actually mean on a body-composition scale?
In this dataset, lean body mass refers to all non-fat tissue: body water, organs, bone, and other non-muscle lean structures — not skeletal muscle alone. Measurements came predominantly from bioelectrical impedance scales; roughly two-thirds of the paired records used Tanita devices.
Why does the headline change so much after adjusting for weight loss?
Unadjusted, the lean-mass gap was −2.1 percentage points. Patients on tirzepatide lost more total weight (−15.2% vs −9.7%), so more absolute lean tissue. After adjusting for concurrent weight change, the gap narrowed to −0.1 pp — roughly 20% of weight lost was lean mass on either agent.
What does the “Depletive” phenotype mean?
Depletive was an exploratory definition combining greater than 20% body-weight loss with greater than 5% lean-mass loss. It occurred in 10.3% of tirzepatide records versus 6.7% of semaglutide records. Within this cohort, the phenotype was concentrated largely among women and adults under 65.
What does the “Prime” phenotype mean?
Prime was the complementary exploratory category: greater than 10% body-weight loss with lean-mass loss under 5%. It was recorded in 11.8% of tirzepatide records versus 12.3% of semaglutide records — P = 0.66, meaning no statistically significant difference emerged between the two groups.
Why are 670,422 screened users and 7,965 paired measurements different populations?
The 670,422 figure counts screened first-episode users; the 7,965 figure counts users with paired baseline and 12-month body-composition measurements — about 1.2%. That paired subset was skewed toward treatment persisters with heavier monitoring, so selection bias remains a stated caveat of the paired analysis.
What did the exercise-intolerance findings actually show?
Greater lean-mass loss was accompanied by exercise intolerance documented 7.2–11.1 percentage points more often. This is an association between chart documentation and measured composition, not a test of strength or performance — no objective exercise testing was performed, and reverse causation is equally compatible.
What does the T-REx randomized trial add?
T-REx randomized tirzepatide participants to structured resistance training or control. Lean soft tissue fell −3.57 kg with training versus −4.67 kg without — a between-group difference of 1.09 kg (95% CI 0.43–1.75) — while weight and fat loss were similar, providing interventional composition data.
Do any of these findings apply to catalog reference materials?
No. The findings describe regulated prescription products used in routine clinical care under physician management. Catalog research reference materials are separately manufactured compounds intended for in-vitro laboratory research — no clinical outcomes, body-composition findings, or care records transfer to them in any way.
Why doesn’t the analysis offer a diet-based explanation for the differences?
The observational analysis did not control diet, protein intake, or resistance training between groups, and those variables were inconsistently captured in the records. Without structured dietary data, no dietary comparison or dietary explanation for the composition differences can be made from this dataset.
Research disclaimer: For research and laboratory use only. Not intended for human or veterinary use. This article does not constitute medical advice. BioPharma provides no dosing, preparation, route, administration, treatment, efficacy, or safety instructions.
