Survodutide sits in an unusual position in this catalogue. It has no labelled dose, because it is approved nowhere — and yet its dose record is more complete than almost anything else here, because the entire dose-finding programme was registered before it was run.
You can watch a dose ladder being built and then cut down.
What the compound is
Survodutide is a 29-amino-acid peptide derived from glucagon, engineered to activate both the glucagon receptor and the GLP-1 receptor. It carries a C18 diacid that binds albumin and prolongs its action — the same protraction strategy the Mounjaro label describes for tirzepatide's C20 diacid and the Ozempic label describes for semaglutide.
Its development name is BI 456906, its sponsor is Boehringer Ingelheim, and its design descends from oxyntomodulin, a natural gut hormone that is a weak dual agonist at the same two receptors.
It is a dual agonist, not a triple one — which distinguishes it from retatrutide, and its action at the glucagon receptor distinguishes it from semaglutide and from the approved GIP and GLP-1 agonist whose dose record sits beside it here, neither of which acts there at all.
Every registered dose
From the ClinicalTrials.gov v2 API on 20 September 2026. All doses are subcutaneous, once weekly; no registered protocol uses any other route or interval.
Phase 2 — the wide ladder
| Trial | Condition | n | Started | Doses tested |
|---|---|---|---|---|
| NCT04667377 | Obesity | 387 | 8 Mar 2021 | 0.6 · 2.4 · 3.6 · 4.8 mg + placebo |
| NCT04771273 | NASH, 48-week dose-ranging | 295 | 27 Apr 2021 | 2.4 · 4.8 · 6.0 mg maintenance + placebo |
Across the two, the tested range spans 0.6 mg to 6.0 mg — a factor of ten.
Phase 3 — the ladder after it was cut
| Trial | Condition | n | Doses carried |
|---|---|---|---|
| NCT06066515 | Obesity | 726 | 3.6 · 6.0 mg |
| NCT06066528 | Obesity with type 2 diabetes | 755 | 3.6 · 6.0 mg |
| NCT06077864 | Cardiovascular safety, event-driven | 5,531 | 3.6 · 6.0 mg |
| NCT06176365 | Obesity, 76 weeks | 274 | 3.6 · 6.0 mg |
| NCT06214741 | Obesity, China | 307 | 3.6 · 4.8 mg |
| NCT06632444 | MASH | 1,800 (recruiting) | not stated by arm |
| NCT06632457 | MASH | 1,590 (recruiting) | not stated by arm |
| NCT07754461 | Type 2 diabetes | 600 (recruiting) | not stated by arm |
Four of the five completed phase 3 trials carry exactly the same two doses. The lowest phase 2 dose, 0.6 mg, does not appear in a single phase 3 protocol.
The finding: the ladder converged, and reading it from the bottom is the error
Set the two phases beside each other and the shape is plain:
| Dose | Phase 2 | Phase 3 |
|---|---|---|
| 0.6 mg | tested | dropped |
| 2.4 mg | tested | dropped |
| 3.6 mg | tested | carried, every trial |
| 4.8 mg | tested | China only |
| 6.0 mg | tested | carried, every trial but China |
This is what dose-finding looks like when it is done in public: a wide spread, then a narrowing to the two figures that survived the trade-off between effect and tolerability.
The practical consequence is that the low numbers in this record are not starting doses — they are rejected doses. A figure of 0.6 mg is real, sourced and registered, and it is a dose the programme tested and then abandoned. A page that presents the range 0.6 to 6.0 mg as "the survodutide dose range" has reported the search space, not the answer, and the difference between the two is the entire value of the phase 3 programme.
What a reader cannot get from this record is the titration — how a trial moved a participant from a first injection to a maintenance dose, which the registry's arm labels describe only as "planned maintenance treatment". That detail lives in the protocols and publications, not in the summary fields, and it is not reconstructed here.
A counting trap, quantified
This site has documented the development-code trap before: adipotide's only human trial is invisible to a registry search for its market name because it is filed as Prohibitin Targeting Peptide 1.
Survodutide gives the same trap a number.
| Search term, 20 Sep 2026 | Studies returned |
|---|---|
survodutide |
26 |
BI 456906 |
31 |
BI456906 |
31 |
Five registered studies of this compound do not surface under its market name. Anyone quoting the size of the evidence base for a development-stage compound should run both searches before quoting a number — and the spacing convention matters too, though here both spellings of the code happen to resolve the same way.
Why the programme's shape is itself informative
Fourteen registered phase 1 studies is a lot, and what is in them says more than the count.
One of them, NCT06352411, is a dedicated renal-impairment pharmacokinetic study: 42 participants sorted into four groups — normal renal function, mild, moderate and severe impairment — started 15 May 2024 and completed. Several others compare formulations in healthy volunteers and in people with obesity.
Studies of that kind exist to answer questions a regulator will ask. They are the most reliable signal in this catalogue that a compound is inside a real development programme rather than adjacent to one — and they are precisely what is missing from the research-stage compounds elsewhere on this site, most of which have a single trial, a terminated trial, or none at all. Compare dihexa, where the registry returns zero studies under any of three names.
It is worth stating the other half plainly: none of this is an approval. A 5,531-participant cardiovascular safety trial that has completed is evidence a sponsor intends to file, not evidence a regulator has agreed. As of 20 September 2026 there is no label for this compound in any jurisdiction, and anything sold under the name today is not a licensed medicine.
Duration
No half-life figure for survodutide appeared in an indexed abstract of the trial programme when this site checked on 20 September 2026. That absence is bounded — it means the abstracts and the registered protocols were searched and the number was not in them, not that it was never measured. The renal-impairment study above certainly measured pharmacokinetics.
What is public is the schedule: every registered protocol, in every phase, gives it once weekly. A weekly interval implies a long half-life, and the albumin-binding C18 diacid explains how it would be achieved, but neither is a measurement and neither is reported here as one. The compound appears on this site's half-life chart in exactly those terms.
Method
The registry census was run against the ClinicalTrials.gov v2 API on 20 September 2026, under both the market name and the development code, taking dose figures from arm-group labels and intervention descriptions. Chemical description is from the published medicinal chemistry literature on the compound.
Dose figures here are registered protocol values, reported as record. They are not established doses, there is no approved dose, and nothing on this page is a dosing instruction.