Cagrilintide Phase 2 Monotherapy Trial: NCT03856047

NCT03856047 was a randomized phase 2 dose-finding trial in 706 adults without diabetes. Five protocol-defined cagrilintide groups had larger modeled body-weight reductions than pooled placebo at 26 weeks. The comparison of the highest studied group with liraglutide nominally favored cagrilintide, but it was not confirmatory. These findings concern the sponsor's test article, selected population, and analysis assumptions; they do not establish long-term health outcomes, FDA approval, or equivalence to separately sourced material.[1][2][3][5]

Cagrilintide is an investigational amylin analogue. This record separates its monotherapy trial from combination studies and from claims about a commercial catalog item.[1][5]

Study at a glance

These design and population fields come from the paper, posted protocol, and current registry.[1][2][5][6] All groups received diet and activity counseling. The group labels below follow increasing protocol-defined study amounts but omit a use regimen; they are not administration instructions.

Randomization, masking, and the analysis question

The manuscript and protocol describe centralized interactive-system randomization. Participants and investigators were masked to active treatment versus matching placebo, but active treatments could be distinguished from one another. The registry's “quadruple” masking classification should not be read as proof that all active comparisons were indistinguishable.[2][5][6]

The primary trial-product estimand asked a hypothetical question: what would the average difference have been if all randomized participants had continued assigned treatment for the full period and had not started another weight-loss intervention? It is not simply the average among people observed while taking treatment. The secondary treatment-policy estimand instead addressed assignment regardless of discontinuation or another weight-loss intervention.[3][5][6]

For the primary analysis, observations outside the treatment-adherent period were treated as missing. The model used analysis of covariance with treatment group and baseline body weight, together with 1,000 within-arm multiple imputations under a missing-at-random assumption. Those assumptions cannot be verified by looking only at the observed weight measurements. The manuscript explicitly cautions that its imputation-based trial-product estimates are not causal effects of the trial product itself.[3][5] The estimands and missing-data guide explains why the target question and the handling of missing observations must be read together.

Primary body-weight result

The following means are the manuscript's modeled trial-product estimates, not the registry's separate descriptive averages. The between-group differences and confidence intervals are the corresponding posted registry statistical comparisons; a negative difference favors cagrilintide over placebo.[2][5]

The registry reports p=0.0002 for group 1 versus placebo and p<0.0001 for the other four cagrilintide groups. The posted statistical analysis plan specifies hierarchical testing from the highest to the lowest cagrilintide group against placebo, controlling family-wise type I error at 5% and stopping at the first nonsignificant comparison. All five reported primary comparisons passed that sequence.[2][3]

The registry explicitly labels its displayed descriptive means as in-trial observations and its statistical comparisons as based on the treatment-adherent period. Its descriptive cagrilintide means are −6.1%, −6.9%, −8.5%, −9.5%, and −10.8%; liraglutide is −8.5% and placebo −3.0%. They must not be mixed with the modeled contrasts as though they were the same summaries. Their differences from the modeled means are not explained merely by rounding.[2][5]

Liraglutide-comparator context and null comparisons

For group 5 versus liraglutide, the modeled difference was −1.81 percentage points (95% CI −3.46 to −0.16; p=0.0316), rounded in the publication to a 1.8-point difference and p=0.03. The posted plan explicitly excludes confirmatory testing of cagrilintide-versus-liraglutide comparisons. This is a nominal result, not multiplicity-controlled proof of superiority.[1][2][3]

Groups 3 and 4 did not show statistically significant differences from liraglutide in these comparisons (p=0.9209 and p=0.3960). The two lower groups nominally favored liraglutide. The highest-group result therefore cannot be generalized to every cagrilintide group, and the null comparisons do not establish equivalence.[2][3] See the guide to negative and null results for that distinction.

Attrition, missingness, and changes to the plan

The paper reports 73 permanent treatment discontinuations and 29 withdrawals from the trial. It reports 30 treatment discontinuations attributed to adverse events across all groups. Study withdrawal, treatment discontinuation, and failure to satisfy the analysis definition of adherence are different events.[1][5]

The manuscript reports 552 participants (78.2%) treatment-adherent through week 26, while 677 attended the week-32 completion visit. Registry completion counts are 97/101, 97/100, 98/102, 101/102, and 94/101 for the cagrilintide groups, 95/99 for liraglutide, and 95/101 for placebo. Separately, the registry's week-26 descriptive outcome counts also total 677 but have a different group distribution. Neither total establishes complete treatment-adherent week-26 data.[2][5]

For the secondary treatment-policy analysis, the plan used retrieved-subject multiple imputation, drawing on observations from participants who stopped treatment but remained in follow-up; this was not another confirmatory testing family. The plan also records changes: the adherence definition was revised because the original was infeasible, a repeated-measures sensitivity analysis was added, and a planned tipping-point analysis was not performed because of computational infeasibility.[3]

The posted plan is version 2, dated March 29, 2020; its external clinical-study-report cover is dated September 18, 2020. The trial had already started, and the registry's primary completion date is March 2, 2020. This review has not established the revision's timing relative to unblinding and does not claim that every method was unchanged from the original plan or that every planned analysis was performed.[2][3]

Adverse events and reporting limits

Gastrointestinal adverse events occurred in 41%–63% of participants across the cagrilintide groups, compared with 32% with placebo. Nausea occurred in 20%–47% and 18%, respectively. Those ranges should not be compressed into an unqualified claim of tolerability.[1][5]

The registry's safety results, labeled week 0 to week 32 and based on all 706 exposed participants, report serious adverse events in 6/101, 2/100, 7/102, 3/102, and 4/101 in the five cagrilintide groups, 4/99 with liraglutide, and 3/101 with placebo. It reports no deaths. These are affected-participant counts, not event counts or proof of causation; the plan treated adverse-event summaries descriptively.[2][3]

The accepted manuscript has an unresolved detail: its adverse-event discontinuation table sums to 29 participants, while its flow diagram and the published abstract give 30. This record retains the published overall count of 30 and does not invent a reconciled by-group breakdown. The manuscript table footnote and registry also label their safety windows differently, so their entire safety datasets should not be assumed identical.[1][2][5]

A trial of this size and duration cannot exclude uncommon, delayed, or population-specific harms. Zero reported deaths is not a general safety conclusion.

Sponsor role and conflicts

Novo Nordisk designed the protocol and statistical analysis plan. Investigators gathered the data; the sponsor provided oversight and collated and analyzed the data. The manuscript reports that all authors had access to data and participated in reviewing the report, with sponsor-funded medical writing assistance under the authors' direction. It lists three authors as sponsor employees and shareholders and discloses other industry relationships.[5]

These roles are part of the evidence record. They neither invalidate the findings automatically nor substitute for independent replication or independent scrutiny of the analysis.

What the findings do and do not establish

The study supports a dose-ranging weight-change signal under its defined trial-product analysis. The placebo-adjusted modeled differences were approximately 3.0–7.8 percentage points. A confidence interval describes uncertainty within that statistical analysis, not every possible bias or every assumption about missing outcomes.[1][2][3][5] The confidence-interval and p-value guide provides additional context.

The 26-week trial did not establish long-term morbidity, mortality, or lasting effects after discontinuation. Its selected population excluded diabetes and women of childbearing potential, and most participants were White. Active-comparator masking was limited, the sources display different site totals, and missing-data assumptions and sponsor involvement remain relevant.[2][5][6]

As checked on September 2, 2026 in Edmonton / September 3 UTC, FDA states that cagrilintide is not a component of an FDA-approved drug, has not been found safe and effective for any condition, and cannot be used in compounding under federal law.[4] Cagrilintide remains an amylin analogue even though this FDA warning appears on a page titled around unapproved GLP-1 drugs.

A trial registration, shared compound name, analytical document, or research-use label does not authenticate another material or establish equivalence to the sponsor's test article. The regulatory-status checking guide and approved-drug versus research-material guide explain those separate questions.

Frequently asked questions

What was the primary endpoint?

Percentage change in body weight from baseline to week 26, evaluated under the defined primary analysis.

How many participants were included?

All 706 randomized participants received study treatment and belonged to the full and safety analysis sets. That does not mean 706 complete treatment-adherent week-26 measurements were observed.[2][5]

Did every cagrilintide group differ from placebo?

Yes, for the reported modeled primary endpoint under the plan's highest-to-lowest hierarchical testing sequence. That conclusion does not extend automatically to every secondary outcome.[2][3]

Was cagrilintide superior to liraglutide?

The highest studied group nominally favored cagrilintide, but the plan specified nonconfirmatory comparator testing. Two intermediate groups had nonsignificant comparisons, and the lower groups nominally favored liraglutide. This trial does not establish a general superiority claim.[2][3]

Why do the paper and registry display different mean changes?

The modeled trial-product estimates and the registry's descriptive in-trial observations use different summaries and observation-period rules. They should not be substituted for one another or explained solely as rounding.[2][5]

Does the study establish long-term safety?

No. Adverse-event summaries from a 26-week treatment trial cannot characterize all uncommon or delayed harms, and a reported event is not automatically caused by the intervention.

Is cagrilintide FDA approved?

FDA's page checked for this review states that it is not a component of an FDA-approved drug and has not been found safe and effective for any condition.[4]

Primary and official sources

1. Published paper record and abstract: PMID 34798060
2. ClinicalTrials.gov: NCT03856047 and posted results
3. Posted statistical analysis plan: version 2, March 29, 2020
4. FDA: concerns with unapproved GLP-1 drugs used for weight loss
5. Author-accepted manuscript: UCL Discovery PDF, identified as an accepted manuscript in the university repository record
6. Posted protocol: version 2, October 2, 2018, with amendments

Review and corrections

Sources checked September 2, 2026 (America/Edmonton) / September 3, 2026 (UTC). This review used the published abstract, the accessible author-accepted main manuscript, relevant posted protocol and statistical-plan sections, current registry results, and FDA's status statement. The separate supplementary appendix was not obtained; the accepted manuscript is not the final publisher version. The current PubMed record links an accompanying commentary, with no erratum, retraction, or expression-of-concern relationship returned in that check. This is not a guarantee that no notice exists elsewhere.

Serum Research is responsible for this editorial summary. No independent scientific, clinical, statistical, analytical, or regulatory review is claimed. This page is not medical advice, a self-use guide, a product recommendation, a material-quality assessment, or a test-article equivalence finding. Send proposed corrections through the public contact form. See how we review evidence, the study-records index, the metabolic research hub, and the Research Library.