Ecnoglutide

XW003

Ecnoglutide (XW003) is a metabolic & fat loss research compound. cAMP-biased long-acting GLP-1 analog — selective cAMP signaling over beta-arrestin, aiming for more efficacy per dose.

How Ecnoglutide works

cAMP-biased long-acting GLP-1 analog — selective cAMP signaling over beta-arrestin, aiming for more efficacy per dose.

Proposed benefits

Researched for fat oxidation, appetite and energy regulation, insulin sensitivity and endurance capacity.

Where to get Ecnoglutide

The evidence for Ecnoglutide

Graded by what exists behind each claim.

✅ Clinically validated

• Phase 3 in China, developed by Sciwind. Trials in obesity and type 2 diabetes have reported weight loss in the mid-teens percent range at higher doses over roughly 48 weeks.
• The trials are Chinese-population, and that changes how the percentages read — lower baseline BMI and different body composition make direct comparison with Western obesity trials unreliable in both directions.

📊 Correlative data

• No Western approval and no independent real-world record. GI tolerability in the trials looks class-typical — nausea early, dose-escalation dependent, easing over weeks.
• It does appear in the research-chemical market, which is worth naming: what is sold there is not the trial material, and none of the phase 3 efficacy or purity assurance carries across to an unverified vial.

🧪 Theoretical / extrapolated

• A GLP-1 analog engineered as a cAMP-biased agonist — it preferentially activates the cAMP signaling arm of the receptor while recruiting less β-arrestin.
• Biased agonism is the interesting idea here. β-arrestin recruitment drives receptor internalization and desensitization, so a cAMP-biased agent should keep the receptor available longer. Whether that translates into a clinically distinguishable advantage is still an open question across the whole GPCR field, not just this drug.

How to read these tiers: they say how much human evidence exists, not how well something works — and ✗ flags harm, never a disappointing trial. How the evidence tiers work →

What Ecnoglutide actually does

This is the one compound in the cohort whose defining number is published, and it is not a receptor selectivity ratio — it is a signaling bias ratio, which is a different and much stranger thing.

In receptor assays, ecnoglutide induced cAMP with an EC50 of 0.018 nM and induced GLP-1 receptor internalization with an EC50 greater than 10 µM Guo 2023. Put those two numbers on the same scale: 0.018 nanomolar against more than 10 micromolar is a separation of more than 500,000-fold between the two things this receptor can be made to do. That is not a tweak. It is a molecule that turns the receptor on and essentially never tells it to leave the membrane.

Why that would matter, stated as the mechanism rather than as the sales pitch. The GLP-1 receptor is a class B1 G-protein-coupled receptor. Agonist binding does two separable things: it couples to Gs and raises cyclic AMP, which is the therapeutic signal, and it recruits β-arrestin, which pulls the receptor off the cell surface into an endosome. Internalized receptor is receptor that is not available to the next molecule of drug. A cAMP-biased agonist is therefore a bet that keeping receptors on the surface produces more signal per unit of drug over a week of continuous exposure than a balanced agonist does — which matters most for a once-weekly drug, because a once-weekly drug is asking the same receptor population to keep responding for seven days.

Now the part that is not on any vendor page: the bias was measured in a cell line, and the clinical claim has never been tested against it. The 500,000-fold number is an in-vitro observation Guo 2023. No human study has measured GLP-1 receptor availability, receptor density, or tachyphylaxis on this drug versus a balanced one. The mechanism is elegant and the mechanism is untested in a person; both of those are true at once and the second one is the one nobody writes down.

The half-life engineering, which is separate from the bias and just as deliberate. Two modifications, each solving a different problem Guo 2023. First, an alanine-to-valine substitution at position 8 — position 8 is where dipeptidyl peptidase-4 cuts native GLP-1 in about two minutes, and a valine there is a bulkier, β-branched residue the protease handles badly. Second, a γ-glutamate–2×AEEA linker carrying a C18 fatty diacid, which binds albumin reversibly: the albumin-bound fraction is too large to be filtered at the glomerulus and is shielded from circulating peptidases, and the free fraction is topped up from that reservoir Muller 2022. Cut the protease site, then hide the molecule inside the largest protein in plasma. The measured result is a steady-state half-life of 124 to 138 hours.

And what it deliberately does not do. One receptor. No GIP arm, no glucagon arm. Everything the duals and triples in this catalog buy with a second and third receptor — the extra energy expenditure, the steeper liver-fat curve — is absent here, and so is the heart-rate cost that comes with it. This is a GLP-1, engineered carefully, and its argument is efficiency rather than reach.

Cell, rodent, human — and where it stops

Step one, in cells. The cAMP and internalization assays above: EC50 0.018 nM for cAMP, > 10 µM for internalization Guo 2023.

Step two, in rodents, and here is a claim worth being precise about. In db/db mice and a diet-induced-obesity rat model, ecnoglutide significantly reduced blood glucose, promoted insulin induction, and produced more pronounced body weight reduction compared to semaglutide Guo 2023. That is a real head-to-head against the market leader — in rodents, at doses chosen by the sponsor, with no human replication. It is also the only place in this compound’s entire record where it beats semaglutide at anything.

Step three, phase 1. Single ascending doses 0.03 to 1.0 mg and multiple ascending doses 0.2 to 0.6 mg once weekly for 6 weeks in healthy participants (NCT04389775). Generally safe and well tolerated; adverse events were decreased appetite, nausea and headache; steady-state half-life 124 to 138 hours Guo 2023.

Step four, phase 2 in type 2 diabetes. 145 adults randomized to 0.4, 0.8 or 1.2 mg weekly or placebo for 20 weeks (CTR20211014). HbA1c fell 1.81%, 1.90% and 2.39% against 0.55% on placebo, P < 0.0001. At 1.2 mg, 71.9% reached HbA1c ≤ 6.5% against 9.1% on placebo, and 33.3% lost ≥ 5% of body weight against 3.0% Zhu 2024. Note the shape: a 2.39% HbA1c drop is very large, and the weight response at the same dose is modest. In this trial it behaved more like a glucose drug than a weight drug.

Step five, phase 3 in obesity, and the doses double. 36 centers in China, 664 participants without diabetes, randomized to ecnoglutide 1.2, 1.8 or 2.4 mg weekly or volume-matched placebo for 40 weeks (NCT05813795). Least-squares mean weight change at week 40: −9.1%, −10.9% and −13.2% against +0.1% on placebo; estimated treatment differences −9.2% (97% CI −11.0 to −7.5), −11.1% (−13.1 to −9.1) and −13.3% (−15.3 to −11.3), all p < 0.0001. At least 5% weight loss was reached by 77%, 84% and 87% against 16% on placebo Ji 2025.

Step six, the head-to-head, which is the most interesting trial here and the least quoted. EECOH-2: 621 people with type 2 diabetes on metformin monotherapy, open-label, 52 weeks, non-inferiority, ecnoglutide 0.6 or 1.2 mg against dulaglutide (NCT05680129). At week 32, HbA1c fell 1.91% on 0.6 mg, 1.89% on 1.2 mg and 1.65% on dulaglutide; treatment differences versus dulaglutide were −0.26% (95% CI −0.39 to −0.13) and −0.24% (−0.38 to −0.11; p = 0.0002). Discontinuations for adverse events were 3%, 4% and 3% He 2025.

Step seven: it is an approved drug. Ecnoglutide received first approval in China in January 2026 for glycemic control in type 2 diabetes and in March 2026 for long-term weight management in adults with obesity or overweight with a weight-related comorbidity, with ongoing evaluation in adolescents and in obesity with obstructive sleep apnea or knee osteoarthritis Shirley 2026. This site’s card still reads ‘Phase 3 human’.

The obstacles, named one at a time. (1) Every efficacy trial is Chinese-population, at a lower baseline BMI than Western obesity trials; the phase 3 enrolled at BMI ≥ 28 or ≥ 24 with a comorbidity Ji 2025, thresholds that would not define obesity in a US trial, so the percentages are not directly comparable in either direction. (2) The one head-to-head is against dulaglutide, and the margin is 0.26 percentage points of HbA1c He 2025. Dulaglutide is the weakest weekly GLP-1 in wide use. A biased-agonist thesis that predicts more signal per dose has been tested exactly once, against the softest available comparator, and won by a quarter of a point. (3) No head-to-head against semaglutide or tirzepatide in humans, so the rodent result that beat semaglutide Guo 2023 stands alone. (4) No cardiovascular outcome trial, which matters because that is the specific thing dulaglutide has and this does not. (5) No published body composition. (6) What is sold in the research-chemical market is not the approved product, and none of the numbers above transfer to an unverified vial.

Ecnoglutide pharmacokinetics — how much of it actually gets in

The catalog says ‘~124–138 hrs’, and for once that is a measured number rather than a schedule. It is the steady-state half-life from the phase 1 study Guo 2023, and it is the only measured half-life anywhere in this cohort. Everything below follows from it.

What the number means in practice. 124 to 138 hours is 5.2 to 5.8 days. Weekly dosing at roughly one half-life per interval means trough concentration sits near half of peak, and accumulation to steady state takes about 4 to 5 doses — a month. Two consequences nobody states: a person judging this drug in the first three weeks is judging a sub-steady-state exposure, and a person who stops still has measurable drug on board for three to four weeks. Side effects do not stop when the injections do.

What holds it in the body. The C18 fatty diacid on a γGlu-2×AEEA linker binds albumin reversibly; bound drug is too large for glomerular filtration and shielded from peptidases Guo 2023 Muller 2022.

What degrades it. The Ala8Val substitution blocks the dipeptidyl peptidase-4 cut that ends native GLP-1 within minutes Guo 2023. What remains is proteolysis of the free fraction by plasma and tissue peptidases and renal handling of whatever is not albumin-bound. It is not a cytochrome substrate, and that has actually been tested rather than assumed: in 28 healthy volunteers, co-administration changed rosuvastatin AUC by a geometric mean ratio of 106% (90% CI 94–120) and digoxin AUC by 84% (76–94), with digoxin Cmax moving from 1.39 to 1.31 ng/mL Li 2026. The digoxin signal is the interesting one and it is not a metabolic interaction at all — it is delayed gastric emptying shifting the absorption of a narrow-therapeutic-index drug by about 16%. That is the real interaction mechanism for this whole class and it applies to every orally dosed drug taken alongside it.

The oral barrier. Absolute for this molecule. An acylated 31-plus-residue peptide meets gastric acid, pancreatic proteases and brush-border peptidases, with hepatic first-pass extraction behind them. Every dose in every trial was injected.

The injectable comparator. Subcutaneous, once weekly, 0.03–1.0 mg in phase 1 Guo 2023 through 2.4 mg in the obesity phase 3 Ji 2025. The site’s dose card reads ‘0.4mg-1.2mg+’, which is the diabetes phase 2 range Zhu 2024; the weight-management doses that produced −13.2% start where that range ends.

What would have to be true, and how you would know it was not

Four predictions. The first is the one that would actually test the biased-agonist thesis, and the third argues against the compound.

1. If cAMP bias does what it claims, tolerance should not develop — and the readout is HbA1c at 32 versus 52 weeks. The argument for low internalization is that receptor stays available, so the same weekly dose keeps working. The EECOH-2 trial ran 52 weeks and reported its head-to-head HbA1c at week 32 He 2025. Draw HbA1c at 3, 6 and 12 months. The prediction is a flat curve after month 3; a drifting-upward HbA1c on a stable dose is receptor desensitization, and it is exactly what this molecule was designed not to do. This is the single measurement that would move the bias claim from a cell line into a person.

2. Fasting insulin should fall alongside HbA1c, not rise. GLP-1 receptor agonism raises insulin secretion in a glucose-dependent way and improves insulin sensitivity through weight loss; the net in trials is better glycemia at lower insulin exposure. Draw fasting insulin with the HbA1c. Falling HbA1c with rising fasting insulin would mean the glucose improvement is being bought with more insulin rather than less, which is not what the mechanism predicts.

3. Lean mass will not be spared, and the phase 3 did not look. More than 25% of total weight lost through pharmacotherapy comes from fat-free mass Stefanakis 2024, and the obesity phase 3 reports no body composition Ji 2025. At −13.2%, that is roughly 3.3% of starting body weight as fat-free mass. Measure grip strength at baseline, 12 and 24 weeks. This is the prediction that cuts against the drug: nothing about signaling bias protects muscle.

4. The lipid panel should improve, and if it does not, look at the diet rather than the drug. The phase 2 in diabetes reported glucose and lipid parameters among its secondary endpoints Zhu 2024. A lipid panel and ApoB at baseline and 24 weeks is the cheapest way to check whether a 10%-plus weight loss is being converted into cardiometabolic benefit, which is the only reason the weight number matters clinically.

What nobody has tested yet

Four experiments, and the first one is the reason this compound exists.

1. Nobody has measured the bias in a human. The 500,000-fold separation between cAMP potency and internalization potency is a cell-line measurement Guo 2023. No study has imaged GLP-1 receptor occupancy, measured receptor density, or tested for tachyphylaxis in a person taking this drug versus a balanced agonist. The entire commercial thesis of this molecule has never been observed in a human. A 26-week crossover of ecnoglutide against semaglutide at matched glycemic effect, with HbA1c and weight measured at fixed intervals, would test it directly and has not been run.

2. Nobody has run it against semaglutide or tirzepatide in people. The only human comparator ever used is dulaglutide He 2025, and the only time ecnoglutide has beaten semaglutide was in rats and mice Guo 2023. That rodent result is the most load-bearing unreplicated claim on this page.

3. Whether the diabetes-versus-obesity split is real. At 1.2 mg in people with diabetes, HbA1c fell 2.39% and only a third lost 5% of body weight Zhu 2024. At the same 1.2 mg in people without diabetes, weight fell 9.1% Ji 2025. Those are different populations in different trials, so the comparison is not clean — but if it holds it means the glucose effect saturates at a lower dose than the weight effect, which would be a genuinely useful dosing principle and nobody has tested it head to head.

4. What is actually in a research-market vial. This is now an approved drug in one country Shirley 2026 and an unapproved research chemical everywhere else. No published analysis of gray-market ecnoglutide exists — no mass spectrum, no purity figure, no confirmation that the Ala8Val substitution or the C18 acylation is present. Without the acylation the molecule is not weekly, and a weekly schedule would deliver two days of drug and five days of nothing.

Ecnoglutide — its own safety story, not its class's

The class block on this page is the shared GLP-1 safety story. Three things are specific to this molecule and belong ahead of it.

1. The tolerability numbers, stated as the trials report them rather than as a summary. In the 40-week obesity phase 3, treatment-emergent adverse events occurred in 93% of participants in each of the three ecnoglutide arms and 84% on placebo; most were mild-to-moderate gastrointestinal events, and ten participants across 499 on drug discontinued for an adverse event Ji 2025. In the 52-week head-to-head, discontinuation for adverse events was 3% at 0.6 mg, 4% at 1.2 mg and 3% on dulaglutide He 2025. Read those together: near-universal side effects, very few of them bad enough to stop for, and no tolerability advantage over dulaglutide.

2. The delayed-gastric-emptying interaction is this drug’s most concrete and most ignored hazard. Digoxin AUC fell to 84% (90% CI 76–94) with co-administration Li 2026. Digoxin has a narrow therapeutic index; a 16% exposure change matters for it in a way it would not for most drugs. The mechanism is not metabolic and generalizes: anything taken by mouth whose effect depends on how fast it is absorbed — oral contraceptives, levothyroxine, warfarin, immediate-release analgesics — is being absorbed on a different schedule while this drug is on board. Nobody warns about that, and it is the interaction most likely to actually happen.

3. Gallbladder disease, sized. Across 76 randomized trials and 103,371 patients, GLP-1 receptor agonist use carried a relative risk of gallbladder or biliary disease of 1.37 (95% CI 1.23–1.52), with a much larger signal in weight-loss trials (2.29, 1.64–3.18) than in diabetes trials (1.27), and higher risk at higher doses and longer duration He 2022. The 2.4 mg weight-management dose sits in the higher-risk stratum on both counts.

4. What the pooled class data does and does not settle. Across 60,080 patients in GLP-1 outcome trials there was no pancreatic cancer signal Sattar 2021. The thyroid C-cell contraindication in medullary thyroid carcinoma and multiple endocrine neoplasia type 2 comes from rodent C-cell tumors, and rodent C cells carry far more GLP-1 receptor than human ones. Calcitonin is a test for people with a family history of those two conditions, not a screening test for anybody else.

5. The honest limit of the record. There is no cardiovascular outcome trial for ecnoglutide. Its own head-to-head comparator, dulaglutide, has one running to 5.4 years. Approval in one country on 40- and 52-week trials Shirley 2026 is a different evidentiary position from a decade of outcome data, and choosing this molecule over dulaglutide is choosing a slightly better HbA1c and a much larger weight effect at the cost of the cardiovascular evidence.

Sources read for this page

• Guo W, et al. Discovery of ecnoglutide - a novel, long-acting, cAMP-biased glucagon-like peptide-1 (GLP-1) analog. Molecular Metabolism 2023 · PMID 37364710
• Zhu D, et al. Efficacy and safety of GLP-1 analog ecnoglutide in adults with type 2 diabetes: a randomized, double-blind, placebo-controlled phase 2 trial. Nature Communications 2024 · PMID 39333121
• Ji L, et al. Efficacy and safety of a biased GLP-1 receptor agonist ecnoglutide in adults with overweight or obesity: a multicentre, randomised, double-blind, placebo-controlled, phase 3 trial. Lancet Diabetes and Endocrinology 2025 · PMID 40555243
• He Y, et al. Efficacy and safety of cAMP-biased GLP-1 receptor agonist ecnoglutide versus dulaglutide in patients with type 2 diabetes and elevated glucose concentrations on metformin monotherapy (EECOH-2): a 52-week, multicentre, open-label, non-inferiority, randomised, phase 3 trial. Lancet Diabetes and Endocrinology 2025 · PMID 40854315
• Shirley M. Ecnoglutide: First Approvals. Drugs 2026 · PMID 42412371
• Li F, et al. Effect of a Novel GLP-1 Analogue Ecnoglutide on the Pharmacokinetics of Rosuvastatin and Digoxin in Healthy Participants. Diabetes, Obesity and Metabolism 2026 · PMID 42310888
• He L, et al. Association of Glucagon-Like Peptide-1 Receptor Agonist Use With Risk of Gallbladder and Biliary Diseases: A Systematic Review and Meta-analysis of Randomized Clinical Trials. JAMA Internal Medicine 2022 · PMID 35344001
• Stefanakis K, et al. The impact of weight loss on fat-free mass, muscle, bone and hematopoiesis health: Implications for emerging pharmacotherapies aiming at fat reduction and lean mass preservation. Metabolism 2024 · PMID 39481534
• Sattar N, et al. Cardiovascular, mortality, and kidney outcomes with GLP-1 receptor agonists in patients with type 2 diabetes: a systematic review and meta-analysis of randomised trials. Lancet Diabetes and Endocrinology 2021 · PMID 34425083
• Muller TD, et al. Anti-obesity drug discovery: advances and challenges. Nature Reviews Drug Discovery 2022 · PMID 34815532

Ecnoglutide — safety, predicted from mechanism

Ecnoglutide — interference & stacking

Bloodwork to run alongside Ecnoglutide

Baseline first, then again at 8–12 weeks.

The Metabolic Health & Prediabetes panel covers these in one order — 8 markers, $81.45 with the discount applied.

Check results you already have → · All 103 markers A–Z

Ecnoglutide — frequently asked questions

What Ecnoglutide is used for

Ecnoglutide appears under 2 goals in the goal router.

Related Metabolic & Fat Loss compounds

← Explore the full Protocol Vault

↑ Back to on this page