Cagrilintide (AM833)

Identity, and where the two registries diverge

AM833 is the one identifier that goes missing. It is a real code, and it titles Fletcher and colleagues' 2021 receptor paper (PMID 33727283), but no registry field carries it; it appears in neither database. Everything else lines up. PubChem holds a single record under the name cagrilintide: CID 171397054, molecular formula C194H312N54O59S2, molecular weight 4409, InChIKey LDERDVMBIYGIOI-IZVMHKDJSA-N. The synonym list on that record adds CAS 1415456-99-3, the development code NN0174-0833, the IUPHAR ligand identifier GTPL13768, and the string 'analogue 23 [PMID: 34288673]', which is the compound number the molecule was assigned in Kruse and colleagues' 2021 paper describing its selection. FDA's Global Substance Registration System holds the matching unique ingredient identifier, AO43BIF1U8, at approved record status, together with the USAN code LM-28 and INN number 11430.

The registration record also publishes the full amino-acid sequence: KCNTATCATQRLAEFLRHSSNNFGPILPPTNVGSNTP, thirty-seven residues, C-terminally amidated. Set against human amylin, KCNTATCATQRLANFLVHSSNNFGAILSSTNVGSNTY, six positions differ. Asparagine 14 becomes glutamate and valine 17 becomes arginine; alanine 25, serine 28 and serine 29 each become proline; and the C-terminal tyrosine 37 becomes proline. Cao and colleagues' 2025 cryo-electron microscopy study, citing the discovery paper for the design rationale, records that N14E and V17R were incorporated to counteract the amyloid fibril formation seen with human amylin, and that the pair was predicted to stabilise the peptide's N-terminal helix through an intramolecular salt bridge. That bridge was seen in a crystal structure of the peptide backbone fused to maltose-binding protein (PDB 7BG0) and again in Cao's own active-state structures.

Lipidation supplies the rest. A C20 fatty diacid is attached at lysine 1 through a gamma-glutamyl spacer, the mechanism by which the molecule is held in circulation long enough for weekly administration. The internal disulfide is stated by GSRS in the peptide's own numbering, linking residues 2 and 7 (sitesShorthand 1_2;1_7), which is the bridge native amylin already carries. PubChem's systematic name gives the same bond as a cyclic (3 to 8) disulfide, a numbering that counts the gamma-glutamyl spacer as residue 1, so the two records agree once the offset is applied. Secondary pages that describe the acyl group as a C18 fatty acid match neither record.

The two registries do not agree on everything. They match on CAS number, UNII, molecular weight and the full 37-residue sequence, and they diverge on one field: PubChem gives the molecular formula as C194H312N54O59S2, while the GSRS property table for AO43BIF1U8 gives C194H312O58N54S2, one oxygen fewer. Computing the formula from the GSRS sequence with the C-terminal amide, the Cys2-Cys7 disulfide, the gamma-glutamyl spacer and the C20 diacid returns C194H312N54O59S2 at 4409.1, which reproduces PubChem's formula and its stated molecular weight. The GSRS oxygen count appears to be the error, and this page uses PubChem's formula on that basis. A two-database check is only worth something if the disagreements are reported alongside the agreements.

A non-selective agonist, by design

Amylin receptors are not single proteins. Each is a heterodimer of the calcitonin receptor with one of three receptor activity-modifying proteins, producing the AMY1R, AMY2R and AMY3R phenotypes, and the calcitonin receptor also reaches the cell surface on its own. Pramlintide, the amylin analogue licensed decades earlier, is selective for the amylin receptors, while cagrilintide activates the isolated calcitonin receptor as well. Fletcher and colleagues profiled AM833 in recombinant cell systems across twenty-five pharmacological endpoints in 2021, comparing it with six other agonists: the lipidated analogues AM1213 and AM1784, pramlintide, salmon calcitonin, human calcitonin and rat amylin. They reported a profile distinct from all six across measures of binding, activation and regulation (PMID 33727283).

Structural work published in 2025 resolved cagrilintide bound to Gs-coupled AMY1R, AMY2R, AMY3R and the calcitonin receptor. Cao and colleagues reported that the peptide adopts an amylin-like binding mode but induces conformational dynamics at these receptors that differ from those induced by rat amylin or salmon calcitonin. The C-terminal proline is one lever: in all three amylin receptors it occupies the same pocket as the tyrosine of rat amylin but makes no direct contact with the modifying proteins, which is consistent with cagrilintide's ability to engage the calcitonin receptor protomer alone.

The same paper tested what the lipid contributes. In a cAMP response element reporter assay, a non-lipidated version of the peptide backbone had similar potency and maximal response to cagrilintide in cells co-expressing the calcitonin receptor and RAMP3, but reduced potency in cells expressing the calcitonin receptor alone. A cagrilintide variant carrying a P37Y substitution behaved the same way: lower potency at the isolated calcitonin receptor, comparable potency once RAMP3 was present. Both results locate the non-selectivity in specific structural features that can be tested one at a time.

Rodent work, and where it points

Carvas and colleagues addressed which receptors carry the effect. Male wild-type and RAMP1/RAMP3 double-knockout littermate mice were fed a high-fat diet for twenty-three weeks, then treated for three weeks with vehicle, salmon calcitonin at 150 nmol/kg, or cagrilintide at 3 nmol/kg, subcutaneously once daily. Wild-type mice given cagrilintide lost 3.4 grams of body weight, dispersion 0.51, at p below 0.005 with eight animals per group, while salmon calcitonin slightly increased it by 0.60, dispersion 0.38. The source does not state whether either dispersion figure is a standard error or a standard deviation. Deleting the two modifying proteins impeded cagrilintide's potency and improved salmon calcitonin's efficacy, which the authors read as dependence on AMY1R and AMY3R.

Jacobsen and colleagues asked a different question in rats: whether weight loss is entirely a matter of eating less. Diet-induced obese male Sprague Dawley rats received subcutaneous cagrilintide, semaglutide or the combination, with pair-fed and weight-matched vehicle controls, ten or eleven animals per group. The combination produced 12 percent weight loss with a 39 percent reduction in food intake, whereas matching that weight loss by restriction alone required a 51 percent cut in intake. On the authors' arithmetic, roughly one third of the combination's effect is attributable to energy expenditure and the remainder to reduced intake.

Ludwig and colleagues built a transcriptomic atlas of more than 530,000 caudal brainstem cells, comprising 80 neuronal populations across rat, mouse and macaque (PMID 42260119). They reported that long-term cagrilintide treatment in rats upregulated prolactin-releasing hormone expression in Calcr/Prlh cells of the nucleus of the solitary tract, and that knocking down Prlh in the dorsal vagal complex abrogated the effects of cagrilintide but not those of semaglutide. Chemogenetic activation of the area postrema Calcr/Ramp3 population, by contrast, failed to affect long-term food intake and body weight.

Cagrilintide alone, in people

Pharmacokinetics come from the 2021 phase 1b trial, in which 95 participants were exposed to treatment, 71 of them to once-weekly cagrilintide 0.16 to 4.5 mg and 24 to matched placebo, all in combination with semaglutide 2.4 mg. Enebo and colleagues reported a cagrilintide half-life of 159 to 195 hours with a median time to maximum concentration of 24 to 72 hours, and found exposure proportional to dose and without effect on semaglutide exposure or elimination. Every published human half-life figure for this molecule therefore comes from co-administration. The renal and hepatic impairment studies reported by Nielsen and colleagues in 2026 did give single doses of cagrilintide on its own, but reported exposure ratios, not a half-life.

Monotherapy efficacy comes from the dose-finding phase 2 trial, NCT03856047, conducted at fifty-seven sites in ten countries. Lau and colleagues randomised 706 adults without diabetes to once-weekly cagrilintide 0.3 to 4.5 mg, once-daily liraglutide 3.0 mg, or volume-matched placebo for twenty-six weeks. The registry's posted participant-flow table gives the arms as 101, 100, 102, 102 and 101 for the five cagrilintide doses, 99 for liraglutide and 101 for pooled placebo. Mean weight reduction on the trial-product estimand ranged from 6.0 to 10.8 percent across cagrilintide doses against 3.0 percent for placebo, and the 4.5 mg group separated from liraglutide by 1.8 percentage points at p equal to 0.03. Gastrointestinal adverse events were reported in 41 to 63 percent of cagrilintide participants against 32 percent on placebo.

The longest monotherapy exposure sits inside a combination trial. REDEFINE 1 randomised 302 participants to cagrilintide 2.4 mg alone for sixty-eight weeks, and Garvey and colleagues reported a mean weight change of minus 11.5 percent on the treatment-policy estimand and minus 11.8 percent on the trial-product estimand. That arm also produced the highest rate of injection-site reactions in the trial: 51 of 302 participants, 16.9 percent, against 12.2 percent for the combination, 2.6 percent for semaglutide alone and 3.0 percent for placebo.

Two further monotherapy arms sit inside combination trials and are rarely quoted. REIMAGINE 2 randomised 152 participants with type 2 diabetes to cagrilintide 2.4 mg alone for sixty-eight weeks; Buse and colleagues reported adverse events in 125 of those 152, 82.2 percent, against 86.9 percent in the combination arm and 70.5 percent on placebo. Earlier, the 32-week phase 2 trial reported by Frias and colleagues gave cagrilintide 2.4 mg alone to 30 participants with type 2 diabetes on metformin with or without an SGLT2 inhibitor, and reported a mean HbA1c change of minus 0.9 percentage points and a mean bodyweight change of minus 8.1 percent, against minus 2.2 points and minus 15.6 percent for the combination.

The combination, and which number gets quoted

REDEFINE 1 randomised 3,417 adults without diabetes in a 21:3:3:7 ratio and ran sixty-eight weeks. Two estimands are reported throughout this programme and they are not interchangeable. The treatment-policy estimand follows participants regardless of adherence; the trial-product estimand models the effect had everyone remained on treatment. Under the first, mean weight change was minus 20.4 percent with cagrilintide-semaglutide against minus 3.0 percent with placebo. Under the second, the same comparison reads minus 22.7 percent against minus 2.3 percent. Both figures are in the paper. The sponsor's regulatory-filing announcement of 18 December 2025 quotes the larger one, which is the figure that travels, usually with no estimand attached.

Results in type 2 diabetes have accumulated quickly. Davies and colleagues reported REDEFINE 2, in which 1,206 participants with type 2 diabetes were randomised 3:1 and the combination produced minus 13.7 percent against minus 3.4 percent for placebo at week 68. In Japan and Taiwan, REDEFINE 5 randomised 331 participants one to one and reported minus 18.4 percent for the combination against minus 11.9 percent for semaglutide alone on the trial-product estimand, 164 and 167 per arm (PMID 42009015). REIMAGINE 2 randomised 2,713 participants across six arms; its primary comparison on the efficacy estimand set the combination at 2.4 mg each, 603 participants, against semaglutide 2.4 mg, 605 participants, and gave an HbA1c difference of 0.16 percentage points, significant at p equal to 0.0035 and small in absolute terms.

One comparison in the programme has no peer-reviewed report at all. REDEFINE 4, NCT06131437, ran eighty-four weeks in 809 participants against tirzepatide 15 mg and is recorded as completed, with no results posted to the registry and no indexed publication located. Its numbers circulate from a company announcement dated 23 February 2026, item id 916501 on the sponsor's press-release archive: 23.0 percent against 25.5 percent under the efficacy estimand, 20.2 percent against 23.6 percent under the treatment-regimen estimand, and a primary endpoint of non-inferiority not achieved. The REIMAGINE 2 report also carries a published correction whose content could not be retrieved.

What the body-composition data showed

Amylin analogues are frequently described in secondary material as sparing lean tissue. REDEFINE 1 contains the only large prespecified measurement bearing on that, and it is a subgroup: 252 of 3,417 participants, 7.4 percent, underwent dual-energy X-ray absorptiometry. In that subgroup, baseline mean total fat mass was 47.4 kg and lean soft-tissue mass 58.1 kg. Absolute decreases from baseline to week 68 with cagrilintide-semaglutide were 17.0 kg of fat and 8.4 kg of lean soft tissue on the trial-product estimand, which the authors state corresponds to 67 percent of total weight loss as fat and 33 percent as lean soft tissue.

Two limits sit on that number. It describes the combination, not cagrilintide alone, so it cannot be used to characterise the amylin component in isolation. And a 7.4 percent scanned subgroup was not powered as a body-composition trial. What the measurement does establish is that a third of the weight lost in this arm was lean soft tissue. No comparable human measurement for cagrilintide monotherapy was located.

Approval status and the shape of the record

Cagrilintide is not an approved medicine anywhere. A company announcement dated 18 December 2025, item id 916470 on the sponsor's press-release archive, records the submission of a New Drug Application to the United States Food and Drug Administration for the fixed-dose combination in weight management, on the strength of REDEFINE 1 and REDEFINE 2, and states that the FDA was expected to review the application in 2026. The February 2026 announcement puts the anticipated decision late in that year. Both announcements describe the type 2 diabetes indication as under investigation in the REIMAGINE programme and neither describes a filing for it. No approval had been announced as of this review, and cagrilintide as a single agent is not the subject of any filing located.

A ClinicalTrials.gov intervention search returned 43 registered studies naming cagrilintide, and the lead sponsor on every one is Novo Nordisk A/S. Two are withdrawn with zero enrolment. That concentration is worth carrying into any reading of the record: the phase 3 dataset is unusually large for this category, and it is also entirely one sponsor's, with the largest head-to-head comparison against a competitor product still unpublished. Independent academic groups appear as authors on the receptor and rodent papers, though several of those papers also carry sponsor employees as co-authors and one was funded by a sponsor consortium grant.

What is not known

The published record establishes little about cagrilintide as an isolated agent over long periods. Its longest monotherapy exposure is the 302-participant, 68-week arm of REDEFINE 1; the second is the 152-participant, 68-week cagrilintide 2.4 mg arm of REIMAGINE 2; the third is a 30-participant, 32-week arm in the 2023 phase 2 trial in type 2 diabetes. Outside those, and outside the 26-week phase 2 dose-finding trial, the molecule has been studied as one half of a fixed-dose combination, so effects attributed to the amylin component are usually inferences from arm-to-arm comparison and not direct measurements of it. Gastric emptying, the property most often asserted for it, has no located primary measurement in humans. Body composition has been measured in one 7.4 percent subgroup, for the combination only. Weight regain after discontinuation is unreported in the published trials. The head-to-head comparison against tirzepatide, the only trial pitting it against a licensed competitor, is completed and unpublished, with no registry results posting. Bone metabolism and muscle health have phase 1 protocols recruiting and no reported endpoints. No cardiovascular outcome data exist: REDEFINE 3, NCT05669755, records 7,101 participants actually enrolled and is recorded as active and not recruiting, with nothing reported. All 43 registered trials share one sponsor, and the discovery paper's preclinical dataset sits behind a paywall this review could not pass, so the rodent and receptor figures quoted here come from other groups' publications.

Questions