Evidence-first notes on bioscience and deep tech, at the edge of the lab and the market. Information only — not investment advice, and not medical advice.
The 30-second version
- What. The GLP-1/GIP headline of −15 to −20% body weight comes with clinical costs: loss of lean (fat-free) mass, real-world discontinuation (roughly 50%+ within a year), rebound after stopping (about two-thirds of the lost weight regained), and safety signals adjudicated by regulators. This piece sizes each cost against the source trials, the evidence tier, and the regulatory verdict (FDA/EMA).
- So what. The costs differ sharply in evidence grade. Lean-mass loss, discontinuation, and rebound are quantified by RCT and observational data; the safety signals split cleanly into those regulators have acknowledged (NAION, gallbladder, ileus) and those they have rejected/de-escalated (suicidal ideation). This is not a hidden catastrophe — but it is not cost-free either.
- Now what. The balance is to neither overstate observational associations as causation nor understate a signal the EMA has acknowledged. Metrics to watch: (1) functional evidence for muscle-preserving combinations (bimagrumab); (2) durability of cardiovascular/renal outcomes after discontinuation; (3) whether the FDA reflects NAION on the label.
The underlying knowledge asset carries a skeptic verdict of proceed-with-caveats (conditional) — neither verified-clean nor hold. It is a conditional go, contingent on the mandatory caveats below.
The five-minute read
GLP-1 discussion usually centers on “how much do you lose (weight %).” This piece flips the question to “at what clinical cost do you lose it.” Behind the −15 to −20% headline sit four categories of cost, each resting on evidence of differing strength.
Lean-mass loss and discontinuation — costs that grow in real-world use
A meaningful share of rapid weight loss is not fat but lean (fat-free) mass. By RCT DEXA substudies, roughly 40% of the weight lost on semaglutide and roughly 25% on tirzepatide is lean mass (this is a cross-trial comparison, not a head-to-head DEXA study). The two-sided point is that fat is lost faster, so body-fat percentage (the ratio) improves. A key limitation: DEXA lean mass is mass, not strength or function. So “lean loss = pathological sarcopenia” is not yet established.
The second cost is discontinuation. One-year discontinuation in RCTs is 17.1% for semaglutide (STEP 1) and 14.2–16.4% for tirzepatide (SURMOUNT-1), but in real-world observational cohorts one-year discontinuation is about 50%+ (46.5% to 64.8% depending on cohort) — roughly 3× the RCT rate. Discontinuation means loss of effect: of those who stopped within 3 months only 3.6% lost weight, and of those stopping at 3–12 months only 6.8% did — a wide gap from the RCT −15 to −20% (an observational association, confounded by titration, cost, and side effects).
Rebound and safety signals — sorted by regulatory verdict
On stopping, about two-thirds of the lost weight is regained within a year (STEP 1 extension), though not a full return to baseline; modeling suggests roughly 25% partial benefit persists as a plateau. This is less a drug flaw than a sign that obesity is a chronic disease — but commercially and for access it points straight to lifelong dosing.
Safety signals must be read with both the evidence tier (RCT hard > observational cohort > pharmacovigilance/label > case report) and the regulatory verdict. What regulators have acknowledged (NAION acknowledged by the EMA as “very rare”; gallbladder and ileus reflected on labels) is clearly separate from what they have rejected/de-escalated (suicidal ideation: FDA requested removal of the warning; EMA found no association). Two traps: NAION was acknowledged, but its size shrank sharply from the original study HR 4.28 to about OR 1.2 in large population studies (“acknowledged” is not “large”). The thyroid C-cell warning is a precautionary boxed warning based on rodent data, with no human confirmation (“warning” is not “confirmed”).
| Cost / signal | Strongest evidence tier | Regulatory status |
|---|---|---|
| Lean-mass loss | RCT DEXA substudy | Mass, not function — not established as pathological |
| Discontinuation | Real-world observational cohort | Association ≠ causation |
| Rebound | RCT extension + modeling | Chronic-disease signature |
| NAION | Observational (single-center → population) | EMA acknowledged (“very rare”); FDA label reflection [unverified] |
| Gallbladder/biliary | RCT meta-analysis | On label |
| Pancreatitis | RCT meta-analysis (no significant association) | Class caution (precautionary) |
| Thyroid C-cell/MTC | Rodent data; no human increase | FDA boxed warning — precautionary, rodent-based |
| Suicidal ideation | Pharmacovigilance signal → review | FDA/EMA rejected — de-escalated |
| Gastroparesis/ileus | Pharmacovigilance + label | FDA acknowledged (ileus added to label) |
Deep dive
1. Background: shifting the question from “how much” to “at what cost”
Parts 0–1 of this series covered efficacy and hard outcomes (SELECT: cardiovascular events −20%; FLOW: renal −24%); Parts 2–3 covered pipeline and commercialization. This part applies a skeptic lens: how much of the headline weight-loss % translates into real-world benefit, and what clinical costs (lean loss, discontinuation, rebound, safety) accrue along the way. One principle governs: do not overstate observational studies as causation, yet do not understate real signals confirmed by labels and regulatory verdicts — a boundary that runs both ways. On the CKM (cardio-renal-metabolic) axis, these costs may be amplified in frail and renally vulnerable populations (see §4, §7).
2. What this asset compiled — the actual size of each cost
Lean-mass loss (RCT DEXA substudy). In the semaglutide STEP 1 substudy, 2.4 mg over 68 weeks produced ~15% weight loss with total fat mass down ~19% and total lean mass down ~10% — about 40% of the weight lost was lean mass. In the tirzepatide SURMOUNT-1 substudy, about 75% of the loss was fat and about 25% lean, directionally favoring lean preservation — but this is a cross-trial, not head-to-head, DEXA comparison. Muscle-preserving combinations have moved beyond hypothesis to demonstration. In BELIEVE (Phase 2b, Nature Medicine 2026, N=507), high-dose bimagrumab plus semaglutide produced 22.1% weight loss of which 92% was fat; lean-mass change was −7.4% for semaglutide alone versus −2.9% for the combination (P<.001) versus +2.5% for bimagrumab alone. Activin type II receptor inhibition quantitatively buffered GLP-1 lean loss (though this is Phase 2b, with side effects including muscle spasm and acne).
Two uncertainties attach. First, a real-world digital-phenotyping preprint (medRxiv 2026) reports that in routine care tirzepatide caused greater lean loss than semaglutide — the opposite direction to the RCT substudies (preprint, non-randomized, different measurement method → flagged as variance). Second, DEXA lean mass is mass, not strength or function, so it falls short of confirming functional sarcopenia.
Discontinuation and durability. As described in the five-minute read, real-world discontinuation runs about 3× the RCT rate; on stopping, about two-thirds of the lost weight is regained (STEP 1 extension: after 17.3% loss, 11.6 percentage points regained over one year off drug), while eClinicalMedicine 2026 modeling suggests a plateau at about 75.3% (95% CI 68.9–81.6), implying roughly 25% partial benefit persists.
3. Methodological strengths and limits — evidence tier decides the conclusion
| Signal | Evidence tier | Regulatory verdict | Actual size / caveat |
|---|---|---|---|
| NAION (optic neuropathy) | Observational (single-center → population). Hathaway 2024 HR 4.28 (95% CI 1.62–11.29, n=710 T2D) → population study OR 1.19 (1.02–1.39, any GLP-1); sema T2D OR 1.24 (0.93–1.66, NS); liraglutide 1.53; Danish cohort (n=424,152) ~5-yr risk doubled | EMA PRAC acknowledged (2025-06): “very rare” (≤1/10,000); product information updated 2025-09-30. FDA label reflection is [unverified] | Relative risk is a signal, but absolute risk is tiny (NAION itself is rare). Original study prone to referral bias and ICD miscoding; effect size collapses in large populations (4.28 → ~1.2) |
| Gallbladder/biliary | RCT meta (76 RCTs) RR 1.37 (95% CI 1.23–1.52), +27/10,000/yr | Cholelithiasis listed on label | High causal plausibility (rapid loss + reduced gallbladder motility). Absolute increase is modest |
| Pancreatitis | RCT meta — no significant association | Class caution on label (precautionary) | Reinterpreted as biliary-mediated; no significant increase in acute pancreatitis in RCTs |
| Thyroid C-cell/MTC | Rodent (dose/duration dependent) → no increase in human cohorts (tirze RR 0.348, p<.001) | FDA boxed warning (contraindicated with MTC/MEN2 history) — but rodent-based, precautionary | Human thyroid C-cell GLP-1R expression is minimal → human relevance unconfirmed |
| Mental health / suicidal ideation | Pharmacovigilance signal (2023) → comprehensive review | FDA/EMA rejected: FDA requested warning removal; EMA “no association” | Actively de-escalated by regulators — causation unconfirmed |
| Gastroparesis/ileus | Pharmacovigilance + label | FDA acknowledged: 2023-09 ileus added to Ozempic label | Absolute risk ~1/1,000 person-years (gastroparesis). Delayed gastric emptying is both mechanism and, when excessive, adverse event; pre-procedure discontinuation advised for aspiration risk under anesthesia |
The strength here is separating, for each signal, the strongest evidence tier from the regulatory verdict, so that “a signal exists” is never confused with “causation is confirmed.” The limits: observational studies (discontinuation rates, population NAION) leave causation unconfirmed; discontinuation rates vary by cohort (47% vs 65%, driven by population, definition, and insurance differences, so no single number should be quoted); and functional metrics for lean-mass loss are absent.
4. Neighbouring domains
Sarcopenia → frailty/kidney (CKM). Lean-mass loss from rapid weight loss raises concern about amplification in frail and CKD-vulnerable populations (associated with falls, hospitalization, dialysis prognosis). But mass ≠ function, so the true burden is uncertain; whether bimagrumab combination matters in this vulnerable group is a key follow-up question. AI/imaging. Body-composition measurement (DEXA, bioimpedance, digital phenotyping) can flip the conclusion (see the direction reversal in the §2 preprint). Muscle biology/gerontology. The activin/myostatin pathway (bimagrumab). Behavioral/adherence modeling. Optimizing titration for GI tolerability.
5. Commercialization and market context (TRL, companies)
Access costs create a vicious cycle: (1) net price / payer attrition drives discontinuation → (2) about two-thirds of lost weight is regained within a year → (3) restarting requires GI titration and cumulative cost. Durability demands lifelong dosing while access is intermittent — a structural mismatch. If Part 1’s hard outcomes (SELECT, FLOW) presuppose continued dosing, the access bottleneck could become the rate-limiting step for population-level CKM outcomes (data on outcome persistence after discontinuation remain uncertain). Policy levers (IRA 2027, Medicare Bridge, and so on) are mitigating variables — meaning payer policy, not pharmacology, may determine whether durability is realized.
Related listed companies include semaglutide (Novo Nordisk, NVO), tirzepatide (Eli Lilly, LLY), bimagrumab-class assets (Lilly, via the Versanis acquisition), and developers of amylin/myostatin pathways. In TRL terms, muscle-preserving combinations are at the Phase 2b demonstration stage (BELIEVE), with a Phase 3 readout as the next gate. This account is a factual summary of regulatory and evidence status, not a buy/sell implication for any specific security.
6. The other side (balance — the counter-case that the costs are overstated)
Because this piece is itself a skeptic lens, for balance we also state the counter-argument that the costs are exaggerated. First, lean-mass loss reduces absolute muscle but improves body-fat percentage; without strength and gait-speed data distinguishing adaptive loss (fitted to a smaller body) from pathological wasting, there is insufficient basis to call it “pathological sarcopenia.” Second, many of the safety signals were reviewed transparently by regulators — suicidal ideation rejected, pancreatitis reinterpreted as biliary-mediated — with no evidence of concealment. Third, NAION was acknowledged by the EMA but its size collapsed in population studies and its absolute risk is tiny, so it does not rise to overturning individual patient decisions. Fourth, rebound can be read not as a drug flaw but as evidence that obesity is a chronic disease. The underlying asset’s skeptic gate being proceed-with-caveats rather than hold reflects exactly this balance.
7. What to watch (falsifiable predictions)
- Muscle preservation: if bimagrumab (Phase 3) or amylin combinations demonstrate lean preservation on functional metrics (strength, gait speed), the “lean loss = pathological sarcopenia” hypothesis is supported; if they preserve only mass with no functional change, “mass ≠ function” is confirmed.
- Durability/outcomes: if cardiovascular/renal hard outcomes vanish along with weight rebound after GLP-1 discontinuation, that indicates “outcomes depend on continued dosing”; if they persist, that suggests a pharmacological residual effect.
- NAION: if the FDA follows the EMA in reflecting NAION on the label, or a large prospective cohort establishes causation, that is “signal acknowledgment expanding”; if population-study OR converges to 1, that confirms a “referral-bias artifact.”
References
- Bikou, A., et al. 2025. “GLP-1 Receptor Agonists and Body Composition” (SURMOUNT-1 body-composition substudy). PMC11965027. link
- Drug Discovery Trends. “GLP-1 Impact on Lean Mass” (STEP 1 lean mass ~40% summary). link
- Heymsfield, S. B., et al. 2026. “Bimagrumab plus Semaglutide (BELIEVE Phase 2b).” Nature Medicine. link
- medRxiv preprint. 2026. “Real-world lean loss via digital phenotyping.” link
- Medscape. 2025. “Real-World Study Finds Over 50% Stop GLP-1s Within 1 Year.” link
- Real-world persistence (obesity clinic). PMC12515774. link
- Hathaway, J. T., et al. 2024. “Risk of NAION in Patients Prescribed Semaglutide.” JAMA Ophthalmology (HR 4.28). link
- Population NAION study (OR 1.19–1.24). JAMA Ophthalmology. link
- EMA PRAC. 2025-06. “PRAC concludes eye condition NAION is a very rare side effect of semaglutide medicines.” link
- FDA. “FDA requests removal of suicidal behavior and ideation warning for GLP-1 RA.” link
- Lee, B. 2023-12-03. “FDA Adds Warning To Ozempic Label About Ileus.” Forbes. link
- Pancreatitis/pancreatic cancer RCT meta. PMC12457091. link
- Gallstone/biliary cohort. United European Gastroenterology Journal. link
- “GLP-1 Thyroid Cancer Black Box Explained” (boxed-warning context). link
Disclosure
This post is for information only. It is not investment advice and not medical advice — treatment decisions must be made with your own clinician. The author holds no position in, and has no direct financial interest in, the listed companies mentioned (Novo Nordisk NVO, Eli Lilly LLY, Amgen AMGN, and others). Much of the efficacy and safety data cited here was generated in trials sponsored by listed pharmaceutical companies; safety signals are attributed to the relevant trials, regulatory documents, authors, or preprints, together with the regulatory verdict (FDA/EMA) and evidence tier (RCT hard / observational cohort / pharmacovigilance-label / rodent). Observational studies represent association, not causation, and relative and absolute risks are stated together. The safety-signal descriptions are a neutral account of evidence and regulatory status, not a negative implication about any specific security.
Mandatory caveats (inherited from the source asset’s skeptic gate): (1) observational studies (discontinuation, population NAION) are not overstated as causation, while EMA-acknowledged NAION and gallbladder signals are not understated; (2) lean-mass loss is mass data, not confirmation of functional sarcopenia; (3) safety signals are stated with absolute risk (relative risk alone would mislead); (4) neutral framing and regulatory verdicts are paired throughout to prevent misreading as negative implications for listed companies’ products.
Skeptic verdict inherited: proceed-with-caveats (conditional). This is a conditional go contingent on the four caveats above — neither verified-clean nor hold. Publication status: HOLD — neutral framing to be confirmed with the Principal given safety-signal implications for listed pharmaceutical companies (charter §Escalation). No publication before Principal approval.
Uthando Nethemba 
Leave a comment