Is an AHI improvement durable if weight is regained?

My laboratory results are from the same laboratory each time, drawn fasting, which I gather matters.

This is one of those things that everyone repeats and nobody derives.

This matters practically, not just academically, because it changes what I would do next.

What is the causal chain, and where does it stop being established?

The short version: substantial AHI reduction in the trials that measured it, with the largest effects in people who lost the most weight.

AHI bands are conventionally five to fifteen for mild, fifteen to thirty for moderate and above thirty for severe. A fall from forty-five to eighteen is a two-band move and still leaves moderate disease.

It helps to be literal here: the correlation between weight change and AHI change in these trials is strong but not deterministic — craniofacial anatomy, muscle tone and sleep position all contribute, and none of them respond to weight loss.

Weight loss by any means is long-established as reducing AHI; the surgical literature has shown this for decades, which is the reason the pharmacological result was expected rather than surprising.

Nothing here is medical advice. Untreated sleep apnoea carries cardiovascular and accident risk that a forum is not equipped to weigh.

Cite the polysomnography endpoint, not the sleepiness questionnaire.

The relevant caution is that improvement is not resolution, and stopping established therapy on the strength of a symptom change is the error to avoid.

Oxygen desaturation index and time below ninety per cent saturation are secondary measures that often move further than AHI, because they weight the severe events rather than counting all events equally.

Home sleep apnoea tests systematically underestimate AHI relative to laboratory polysomnography, so a before-and-after comparison should use the same modality or the difference is partly instrumental.

Adherence to positive airway pressure is the main confounder in any real-world comparison and is rarely measured well outside a trial.

Symptom improvement is a poor proxy for index improvement, in both directions.

Get a baseline study before you need a comparison, because you cannot obtain one retrospectively.

The mechanism is largely weight-mediated: reducing upper-airway and abdominal adiposity reduces collapsibility and improves respiratory drive mechanics. That makes the effect real and not specific.

Central apnoeas are counted in the same index but arise from respiratory control rather than airway collapse, and there is no reason to expect a weight-mediated intervention to address them.

Mechanically, daytime somnolence scores improve alongside AHI in these trials, but they also improve with placebo, which is why the instrumented endpoint is the one that carries the argument.

The trials studied licensed product at titrated doses in a monitored population, which is not what an unverified research compound is.

Compare like with like: same modality, same scoring rules, or the difference is partly an artefact.

edited 4 Sept 2025 by micron22 — added the method parameters

The honest answer is that the effect is large in relative terms and that a large relative reduction from a severe baseline can still leave a treatable condition.

Positional dependence matters: an index averaged across the night conceals whether the events are concentrated supine, which changes what a partial improvement is worth.

SURMOUNT-OSA is the dedicated trial in this indication and used polysomnography rather than symptom scales, which is why its result is quotable.

This is one of the few places where the instrumented evidence in this class is unusually clean. Use it.

This is one of the better-evidenced non-glycaemic outcomes in the class, because it was studied with a hard, instrumented endpoint rather than a questionnaire.

SURMOUNT-OSA studied tirzepatide in adults with obesity and moderate-to-severe obstructive sleep apnoea, both with and without positive-airway-pressure therapy, and reported reductions in the apnoea–hypopnoea index of roughly twenty-five to thirty events per hour against a small placebo change.

A two-band improvement is a real result and is not the same as resolution.