Accepted answer
The "miss" was entirely relative to a guidance number, not to any comparator, and the mechanistic explanation you have heard about dose adjustment is substantially correct. But the decomposition arithmetic is the more interesting part of the trial, so let me do that too.
The numbers
REDEFINE 1 randomised roughly 3400 adults with overweight or obesity and without diabetes across 68 weeks, with arms for CagriSema, semaglutide 2.4 mg alone, cagrilintide 2.4 mg alone, and placebo. Approximate mean weight changes: CagriSema about -22.7%, semaglutide about -16.1%, cagrilintide about -11.8%, placebo about -2.3% [1]. The companion trial in type 2 diabetes reported a smaller magnitude, in the region of -13.7% against about -3.4% for placebo, which is the usual diabetes penalty [2].
Decomposing the combination
Work in placebo-adjusted terms, since that is the only way the components add meaningfully:
- Semaglutide alone, placebo-adjusted: 16.1 - 2.3 = 13.8 percentage points.
- Cagrilintide alone, placebo-adjusted: 11.8 - 2.3 = 9.5 percentage points.
- Naive sum if fully additive: 13.8 + 9.5 = 23.3 percentage points.
- CagriSema observed, placebo-adjusted: 22.7 - 2.3 = 20.4 percentage points.
- Observed as a fraction of the additive prediction: 20.4 / 23.3 = 0.876, so about 88% of additivity.
That is a genuinely notable finding and it is not what most combination pharmacology looks like. Near-additivity implies the two components are working through substantially non-overlapping mechanisms. Compare with the strong sub-additivity seen when you add intensive lifestyle intervention to semaglutide, where the increment is a fraction of what the intervention achieves alone. Amylin analogue plus GLP-1 agonist behaves much more like two independent levers.
Caveat on that arithmetic: it uses arm means from a single trial rather than a formal interaction test, and the monotherapy arms were smaller than the combination arm. Treat 88% as an approximation with real uncertainty, not a measured interaction coefficient. It is enough to support "close to additive" and not enough to support a specific number.
Why 22.7% read as a shortfall
Three contributing reasons, in descending order of substance:
- Dose adjustment was permitted and widely used. The protocol allowed investigators to reduce or hold the dose of either component for tolerability, and only a bit over half of participants were on the full 2.4/2.4 mg combination at the end. So the arm mean is a mean over a mixture of doses, and it is closer to a maximum-tolerated-dose result than to a fixed-dose result. That systematically pulls the mean below what a fully escalated arm would have produced, and it is why the phase 2 signal did not fully carry through.
- The expectation was anchored on phase 2 and on informal guidance. Phase 2 numbers in this field come down in phase 3 essentially always, for population, estimand and site-quality reasons. An expectation of 25% built on phase 2 was never a well-calibrated prediction.
- Comparator context. Against tirzepatide's -20.9% in SURMOUNT-1 and -20.2% in SURMOUNT-5, a -22.7% is a small indirect margin that would not survive the cross-trial caveats. So the commercial story of a decisive step change did not materialise, even though the number is the highest published.
The scientific reading and the commercial reading diverge here, and both are internally consistent. Scientifically: two mechanisms combine near-additively, and the combination produced the largest published phase 3 mean weight loss. Commercially: it did not clear tirzepatide by a margin that indirect comparison could defend, and the dose-adjustment design makes the ceiling uncertain.
What the amylin component contributes mechanistically
Amylin is co-secreted with insulin from beta cells and acts at calcitonin and amylin receptor complexes in the area postrema and related hindbrain regions, slowing gastric emptying and producing meal-terminating satiation. Crucially it also appears to influence the defended body-weight set point in a way distinct from GLP-1 signalling, and there is a body of work suggesting amylin agonism can restore leptin responsiveness. The near-additivity in the arithmetic above is the clinical fingerprint of that mechanistic separation.
2The 88%-of-additivity calculation is the most useful thing to come out of REDEFINE and almost nobody ran it. – RP_C18 9 months ago 3Only just over half the arm on full dose is the detail that reframes the whole "miss" narrative. – a_lindgren 37 days ago add a comment