What the GIP receptor actually does.

Almost every popular account of tirzepatide describes the GIP receptor the same way: the second receptor, the bonus mechanism, the thing that makes Mounjaro bigger than Ozempic. That framing is not false. It is pointed at the wrong variable.

Here is the thesis. GIP's most consequential contribution to this drug class is probably not how much weight the receptor removes. It is how much of somebody else's signal the brainstem will accept before a patient stops taking the drug. GIP looks less like an efficacy lever with a tolerability side benefit, and more like a tolerability lever whose efficacy is substantially borrowed.

I think that reframing does real work, because it dissolves the field's most embarrassing open problem. Agonizing the GIP receptor produces roughly 20% weight loss. Antagonizing the GIP receptor also produces roughly 20% weight loss. Read as a question about appetite, that is a paradox. Read as a question about dosing headroom, it is close to expected.

The minimum background. GIP is one of the two incretin hormones, and in healthy people it accounts for the larger share of the incretin effect, though its insulinotropic action is markedly blunted in type 2 diabetes. For two decades the obvious drug was a blocker: a 2002 Nature Medicine paper showed GIP-receptor knockout mice were protected against diet-induced obesity. Lilly then built the opposite of that and won.

The field is publishing both answers in the same issue

If you want one artifact that captures how unresolved this is, take the August 2025 volume of Diabetes. Page 1326: "A Contemporary Rationale for Agonism of the GIP Receptor in the Treatment of Obesity." Page 1334: "GIP Receptor Antagonists in the Pharmacotherapy of Obesity." Same journal, same issue, opposite directions, neither refuted.

The clinical numbers are just as uncooperative. In SURMOUNT-5, the head-to-head at 72 weeks, tirzepatide produced 20.2% mean weight reduction against 13.7% for semaglutide 2.4 mg. GIPR agonism, on top of GLP-1, with a 6.5-point margin.

Then Amgen's maridebart cafraglutide, MariTide, a bispecific that agonizes GLP-1R and antagonizes GIPR, reported up to roughly 20% mean weight loss at 52 weeks in its Phase 2 obesity cohort, with no plateau at week 52, and up to about 17% plus an HbA1c reduction of up to 2.2 percentage points in the type 2 diabetes cohort. Monthly dosing, opposite pharmacology on the GIP arm, similar magnitude.

The honest position in mid-2026 is that no published mechanistic account predicts both of those results from first principles. Everything below is my attempt at the least bad one.

The most underrated finding in this class is an antiemetic one

In 2021, Diabetes published work showing that GIP receptor agonism attenuates GLP-1 receptor agonist-induced nausea and emesis in preclinical models. The anatomy is the interesting part. GIPR turns out to be expressed on GABAergic — that is, inhibitory — neurons in the area postrema and nucleus tractus solitarius, the hindbrain region that both processes emetic stimuli and is required for GLP-1's suppression of food intake.

The effect held across three species, including musk shrews, one of the few laboratory animals that can actually vomit. GIPR signalling blocked emesis and reduced illness behaviour while food intake suppression, weight loss and glucose tolerance were preserved.

A 2025 paper in Science Advances pushed the same finding further: tirzepatide produced hypophagia and body weight loss accompanied by fewer gastrointestinal adverse events than equipotent semaglutide in preclinical models. Not less appetite suppression with fewer side effects. The same appetite suppression with fewer side effects.

Tolerability is efficacy when the ceiling is discontinuation

This is where the mechanism stops being academic.

The binding constraint on this drug class in the real world has never been the size of the effect in a trial. It is that a large share of people stop within the first year, and gastrointestinal intolerance is one of the two reasons that reliably shows up alongside cost. We went through those numbers in the discontinuation data the press releases skipped.

So consider what a mechanism that suppresses nausea is worth on a population basis. It buys titration headroom. It converts a dose a patient theoretically could receive into a dose they actually stay on. It produces weight loss without doing any weight-lowering work itself.

Which is why I think the tirzepatide-versus-semaglutide comparison has been read slightly wrong for three years, including by us. Some unknown share of that 6.5-point gap in SURMOUNT-5 is not a bigger effect. It is a longer runway. We have been attributing to appetite what may belong to the vomiting reflex.

Desensitization is the least glamorous explanation and probably the right one

The paradox still needs an answer on the peripheral side, and the best available one is boring.

In 2020, an Amgen group published mechanistic work in Nature Communications showing that chronic GIP receptor agonism desensitizes adipocyte GIPR — a functional downregulation that ends up resembling antagonism. Sustained agonism, in other words, is a slow route to a partially silenced receptor.

A 2026 comparison in Diabetes, Obesity and Metabolism tested the two directions side by side in male mice. Both a GIPR agonist and a GIPR antagonist reduced appetite and body weight. The agonist was most effective acutely; the antagonist's anorectic effect was sustained across all 17 days of daily dosing. That temporal signature — agonist front-loaded, antagonist flat — is precisely what receptor desensitization predicts.

So the two drug classes plausibly converge on a similar peripheral end state by opposite routes. But they do not arrive with the same central consequences, because only one of them is putting an inhibitory signal into the area postrema on the way. If that is right, the agonist and the antagonist should differ most in adverse-event profile and in the shape of the response curve, and least in final weight.

The strongest case that I am wrong

Two objections, and the first is serious.

One: you can reach this efficacy without touching GIPR at all. CagriSema, cagrilintide plus semaglutide, reported 22.7% at 68 weeks in REDEFINE-1 using amylin, not GIP. Retatrutide reported 24.2% at 48 weeks in Phase 2, where glucagon agonism is doing a large share of the work. If 22 to 24% is reachable through three different receptor combinations, then GIP is not the story, total exposure and complementary appetite pathways are, and my tolerability thesis is an elaborate way of saying some molecules titrate better than others.

I concede a good deal of that. Preclinical antiemetic data in shrews is not human tolerability data. No trial has isolated GIPR's antiemetic contribution in humans, and until one does, this remains a mechanistic story with circumstantial clinical support. I hold the position anyway for one reason: a framework in which GIP is irrelevant still owes an explanation for why the antagonist works, and does not have one.

Two: MariTide's own Phase 2 reported nausea and vomiting as its most common adverse events, which sits awkwardly beside any claim that GIP modulation is fundamentally about emesis. Fair. Though the reported pattern — predominantly mild, transient, concentrated around the first dose, and substantially reduced by dose escalation — is a tolerability curve, not a refutation of one.

A necessary boundary: everything above is mechanism reporting and editorial analysis. Ozemback does not recommend, rank, endorse or discourage any medication, and nothing here describes what any reader should take, start, stop or change. Trial doses are cited as trial design, not as guidance.

What would actually settle this

The test is already running, and it is not a weight curve.

Watch whether adverse-event-driven discontinuation separates between the GIPR-agonist and GIPR-antagonist programs at matched weight loss as the MariTide Phase 3 program reports out. If the antagonist matches on kilograms but not on drop-out, the tolerability thesis survives and GIP's real job becomes visible. If the two look alike on both, GIP is doing less than I think and the field can go back to arguing about exposure.

Either way, the receptor was underestimated — just not in the direction the coverage assumes. It was never the second weight-loss engine sitting beside GLP-1. It looks like the throttle on the first one. That is a smaller job than the marketing implies, and a more important one than the science press has yet priced in.

Ozemback — August 2026