A Biomarker Is Not an Outcome

Researchers measured sixteen cardiovascular risk markers in SURMOUNT-1 participants at 72 weeks, and most of them improved substantially: inflammation down by more than half at the top dose, insulin resistance down 39%, the protective hormone adiponectin up 48%.^[1] It is a striking panel. It is also not evidence that anyone avoided a heart attack, and the difference between those two statements has a body count behind it.

What was measured

Plasma samples were drawn at baseline, 24 weeks and 72 weeks from 100 randomly selected participants in each SURMOUNT-1 arm who completed treatment — placebo, or tirzepatide at 5, 10 or 15 mg — leaving 392 people after samples with insufficient volume were excluded. Sixteen markers were assayed across four categories: inflammation, metabolic and hepatic stress, endothelial function, and clotting.

Adiponectin rising is the right direction — it is the one marker here where more is better. Everything else falling is the right direction too. Fibrinogen, tissue plasminogen activator antigen and thrombomodulin showed no consistent association and appear in no row above, which is worth noticing: three of the sixteen markers did not move.

The rows that spoil a clean story

Most of the table shows an orderly dose gradient — more drug, more effect — which is the pattern that makes a result believable. Two things break it.

PAI-1 antigen runs −41.4%, −35.6%, −44.3% across the three doses. The middle dose performed worst. Nothing about the biology predicts that, and it is the signature of sampling noise in a hundred-person cell rather than a dose relationship.

And four markers were not significant at the lower doses at all — leukocytes, free fatty acids, sICAM-1 and platelets. Free fatty acids and platelets reached significance only at 15 mg. A summary saying “tirzepatide improved sixteen cardiovascular biomarkers” would be describing the top dose and quietly generalizing.

Why a marker is not an event

A biomarker is a stand-in for the thing you care about, and the substitution is only as good as the assumption that moving the marker moves the outcome. Sometimes it does. Sometimes the drug moves the marker by a route that has nothing to do with the disease, or does something else harmful at the same time.

The marker moving is a reason to run the outcome trial. It is not a substitute for having run it.

This does not put tirzepatide in the same position as those examples. Its dose gradients are orderly, the markers point the same way across four unrelated biological systems, and the mechanism is not exotic. All of that raises the odds that the panel reflects something real. It is a stronger prior, not a result.

The sample is only people who finished

Participants were drawn from those who completed treatment. Everyone who stopped — through side effects, through anything else — contributed no sample. That selection is invisible in the numbers and it runs in a predictable direction: people who tolerate a drug well are not a random subset of people prescribed it.

So the panel describes what happens in the blood of someone who took tirzepatide for 72 weeks successfully. That is a legitimate question and a narrower one than it appears.

The outcome trial that would settle the question is running. Until it reports, the honest summary is that tirzepatide moves a wide range of cardiovascular risk markers in the favorable direction, and that nobody yet knows what that buys in events. For a case where the events were counted, see how many people to treat to help one.