What does the mechanism of tirzepatide predict that SURMOUNT-2 did not test?

Numbers first: tirzepatide · SURMOUNT-2.

I suspect the usual explanation for this is wrong, or at least incomplete.

I am aware this may have a boring answer. I would still like the boring answer stated clearly.

So what is the mechanism, and how well established is it?

Read SURMOUNT-2 by arm, because the arm is the unit of randomisation and every figure worth quoting is defined at that level. A programme that randomised several dose levels reports each one separately, with its own sample size and its own interval, and the pooled number that circulates afterwards describes a group nobody was assigned to. Take the primary publication and its supplementary tables rather than a summary of them: one is organised by arm, the other by whichever figure was largest. And check the estimand — what happened to everyone assigned, or what happens to those who kept taking it — because the two answer different questions and are routinely quoted as though they were one.

Answer first: the receptor is a class B G-protein-coupled receptor signalling mainly through Gs and cyclic AMP, and almost every downstream effect people ask about traces back to where that receptor is expressed rather than to what it does when activated.

Glucose-dependence arises because the insulinotropic signal amplifies glucose-stimulated secretion rather than initiating secretion. With no glucose signal to amplify, there is little to amplify.

What each test answers

To be exact about it, gastric emptying delay attenuates with continued exposure for long-acting agents through receptor desensitisation, which is why the early nausea usually settles while the appetite effect persists.

The incretin effect itself was established by comparing the insulin response to oral and intravenous glucose loads matched for plasma glucose; the difference is what the gut hormones contribute.

The caveat is that mechanism predicts direction and rarely predicts magnitude in an individual, and people reason from mechanism to dose far too confidently.

Mechanism is a good guide to what to expect and a poor guide to how much.

This is answerable mechanistically, and the mechanism actually predicts the side-effect profile, which is unusual and worth exploiting when reasoning about it.

Native GLP-1 has a circulating half-life of one to two minutes because dipeptidyl peptidase-4 cleaves the two N-terminal residues. Substituting the position-8 alanine, as the long-acting analogues do, blocks that cleavage and is the single most consequential modification in the class.

It helps to be literal here: biased agonism — differential recruitment of beta-arrestin versus G-protein signalling — is an active research area and is one hypothesis for why agents with similar receptor affinity have different tolerability.

Structural work on the receptor by cryo-electron microscopy has resolved the agonist-bound active state and is the basis for current structure-guided design in this class.

Research-use material is not approved for human use, and mechanism is not a safety argument.

Glucose-dependence is the property to remember; it explains the safety profile on its own.

The honest answer is that appetite suppression is centrally mediated and gastric emptying is peripheral, and the two contribute different amounts in different people.

Glucagon suppression is also glucose-dependent and is lost during hypoglycaemia, which preserves the counter-regulatory response — a genuinely elegant piece of physiology and the reason the class is safe in this respect.

Receptor density and downstream coupling differ between tissues, so the dose-response curves for glycaemia, weight and nausea are not the same curve. That is the pharmacological basis for titration.

A receptor being expressed in a tissue does not establish that activating it there matters at therapeutic exposures.

Tissue distribution first, then signalling. Nearly every question in this tag resolves at the first step.

edited 22 Jun 2024 by Dr_Elias_Weiss — added the placebo-arm figures

On the detail: the endogenous hormone is degraded by dipeptidyl peptidase-4 within a couple of minutes. Every long-acting agent in this class is essentially an answer to that one problem.

Central effects reach the arcuate nucleus and the area postrema, regions with an incomplete blood-brain barrier. That anatomy is why a large peptide can act centrally at all, and it also explains the nausea, since the area postrema is the chemoreceptor trigger zone.

The position-8 substitution conferring DPP-4 resistance appears in essentially every long-acting agent in the class, which is about as strong a piece of convergent evidence as medicinal chemistry offers.

Nothing here is medical advice; this is pharmacology.

Nausea and appetite share an anatomy, which is why they are hard to separate by dose.

The short version: glucose-dependent insulinotropic action, glucagon suppression, delayed gastric emptying and central appetite effects, from one receptor in four places.

Albumin binding does the rest of the work. A fatty-acid chain attached through a linker binds circulating albumin reversibly, which both shields the peptide from renal clearance and creates a depot; that is how a two-minute hormone becomes a once-weekly drug.

The caveat is that mechanism explains and does not predict. A clean mechanistic story has repeatedly failed to survive a Phase 3 in metabolic medicine.

The half-life problem and the albumin-binding solution are the whole story of the class chemically.

edited 22 Jun 2024 by Dr_Bram_Verhoeven — fixed an arithmetic slip in the third paragraph