A study out this week tries to answer a question a lot of people with obstructive sleep apnea and obesity are quietly asking their care team: if I have to pick one lever to pull — a GLP-1, an SGLT2 inhibitor, or just exercising more — which one actually moves the needle on my apnea?
The answer the researchers landed on has a clear headline. It also comes with a list of caveats long enough that the authors themselves warn against using it to make a treatment decision. Both parts matter.
What the study did
Researchers based at Henan University of Chinese Medicine, publishing August 17 in Frontiers in Endocrinology, ran a network meta-analysis — a statistical method for comparing treatments that have mostly never been tested against each other directly — pooling 15 studies (13 randomized trials, 2 case-control) and 1,877 participants across nine countries.
They compared reductions in the apnea-hypopnea index (AHI), the count of breathing interruptions per hour used to grade OSA severity, across three interventions: GLP-1 receptor agonists, SGLT2 inhibitors (a class of diabetes medication), and aerobic exercise programs. All three beat placebo:
- GLP-1 receptor agonists — AHI reduction of 15.28 events/hour
- SGLT2 inhibitors — AHI reduction of 13.68 events/hour
- Aerobic exercise — AHI reduction of 12.50 events/hour
GLP-1 medications also ranked first for BMI reduction and blood oxygen improvement. Aerobic exercise ranked highest on one measure: daytime sleepiness scores.
What “ranked highest” doesn’t mean
This is where the study’s own authors get careful, and where a lot of coverage elsewhere won’t.
The interventions were barely ever tested against each other. A network meta-analysis fills that gap with indirect comparison — inferring how GLP-1s stack up against exercise by comparing each to placebo separately, then triangulating. That’s a legitimate statistical technique, but it’s not the same as a trial that randomized the same population to one arm or the other. The paper says so directly: comparisons between the active treatments were “largely non-significant” once you account for that indirect path.
The confidence in the ranking is low. The authors rate no comparison in the entire analysis as high-certainty evidence. Most sit at low or very low certainty, and a funnel-plot check turned up signs of possible reporting bias — the statistical fingerprint of smaller, less favorable trials being less likely to get published in the first place.
The exercise arm is a grab-bag. “Aerobic exercise” covered a wide range of modality, intensity, and session length across the pooled trials, which is its own source of noise. A person’s real-world result from “start exercising more” depends heavily on which version of that they actually do.
The authors’ own conclusion is unambiguous: the treatment hierarchy should be “regarded as hypothesis-generating rather than as a basis for firm clinical recommendations.” That sentence is doing a lot of work, and it’s worth reading twice before anyone turns this into a pitch that GLP-1s beat exercise for sleep apnea.
The part that applies to compounded medications
The GLP-1 trials folded into this analysis studied branded, FDA-approved products under their approved indications.
Compounded semaglutide and tirzepatide — what NoBsRx provides provider-guided access to — are prepared by state-licensed 503A pharmacies for an individual patient. They are not FDA-approved products, and no arm of this meta-analysis, or any trial feeding into it, tested a compounded preparation for sleep apnea. The evidence here belongs to the drug class and the branded product; it does not automatically transfer to a compounded version of it, and nobody should market it as if it does.
Why the ranking matters less than your own number
Even a clean, high-certainty version of this study would still describe an average across a trial population — not a prediction for any one person. AHI response to any of these interventions varies a lot person to person, and none of these trials can tell you in advance which side of the average you’d land on.
The more useful starting point isn’t “which treatment wins in a pooled analysis.” It’s knowing your own AHI in the first place. Roughly three in four adults with obesity have obstructive sleep apnea, and most have never been tested for it — which means most people asking “GLP-1, SGLT2, or exercise?” are asking without a baseline number to track against.
That’s the gap our sleep program is built around: a 3–5 night at-home study on an FDA-cleared ring, scored by a board-certified sleep clinician, that gives you an actual AHI reading instead of an average borrowed from a study population. Whatever you and your licensed provider decide to do about sleep apnea, it’s a better decision made from your own data than from a ranking in a meta-analysis.
This post is general information, not medical advice. It does not describe a benefit of any NoBsRx program and is not a claim that any medication or exercise program will treat sleep apnea. Compounded medications are not FDA-approved. Treatment is never guaranteed — eligibility, diagnosis, and every treatment decision are made by independent licensed providers based on an individual clinical assessment, and availability varies by state. Do not start, stop, or change any treatment based on an article. If you think you may have sleep apnea, talk to a licensed clinician.