Is It Just the Weight Loss?

Every benefit reported for these drugs runs into the same question: is this the medicine, or is it just the weight coming off? A prespecified analysis of the SURMOUNT-OSA trials tried to separate the two using mediation modeling, and concluded that treating both the sleep-disordered breathing and the obesity is likely needed to get the full cardiometabolic benefit.^[1] The reasoning is worth understanding, and so are its limits.

Why the question is hard

When someone with obstructive sleep apnea takes tirzepatide, several things change at once. They lose weight. Their apnea-hypopnea index falls. Their overnight oxygen levels improve. And their inflammatory markers, insulin resistance and blood pressure all move in the right direction.

Establishing which change drove which is not something a trial answers by design, because the trial only randomized the drug — not the weight loss, and not the apnea improvement. Mediation analysis is the statistical attempt to decompose it after the fact: given how much each intermediate changed, how much of the final outcome moves with it?

What the analysis found

Across both 52-week studies in the SURMOUNT-OSA program, tirzepatide improved cardiometabolic risk factors more than placebo. The mediation results then split into three distinct patterns.

The first three rows are the interesting ones. Inflammation, insulin resistance and triglycerides showed an independent mediation effect from the sleep apnea measures — meaning the model attributes part of their improvement to breathing better at night, over and above what the weight loss accounts for.

That is an argument against “it is all just the weight loss” — and it is a modeling argument, not a demonstration. ★ A far larger prespecified analysis has since reached the same conclusion for the heart, in 17,604 people against a hard endpoint: losing weight predicted more heart attacks.

What mediation analysis can and cannot show

Two further cautions belong with that. Parts of this work were post hoc, including the insulin-resistance analysis and the mediation analysis itself — the paper says so. And the published abstract reports which effects were significant without reporting how much of each outcome was mediated, so we cannot tell you whether sleep apnea accounted for a large share of the inflammation improvement or a small one. We are not going to estimate it.

The author disclosures on this paper are dense with industry relationships, Eli Lilly among them, as declared in the publication. For a manufacturer-sponsored analysis of a manufacturer-sponsored trial, that is the expected picture rather than a surprising one, and it is context a reader should have.

The blood pressure result nobody has explained

Systolic pressure was mediated by weight, and by weight combined with the sleep apnea measures. Diastolic pressure was mediated by neither — not weight, not apnea metrics, not the combination.

Diastolic pressure did not fail to change; it simply was not explained by either intermediate the analysis examined. That leaves an open question rather than a finding: something about the treatment moved it that this decomposition did not capture. We would rather report that as unresolved than construct a mechanism for it.

What it means practically

• Sleep apnea is worth treating on its own terms, not only as something weight loss will eventually fix. If the modeling is right, part of the metabolic benefit runs through the breathing.
• Stopping CPAP because the weight came off may forfeit something. This analysis gives a reason to think the two interventions are not interchangeable, though it does not test that directly.
• “It is just the weight loss” is too simple a story — and so is any confident claim about exactly how much is not.