Skip to content
This GLP
← Research
Evidence

Sleep apnea improved in every group, by very different amounts

Age, sex, BMI and neck size made little difference to the result. Starting severity made a fourfold one, for arithmetic reasons worth understanding.

Glenn Torres6 min read
Fewer breathing events per hour, by subgroupby age27.734.1by sex19.832.6by BMI25.234.4by neck size23.930.8by starting severity12.152.2The bottom row is four times as wide as the others.

Tirzepatide is approved for obstructive sleep apnea with obesity, and the obvious next question is whether it works as well for everyone. Post hoc analyses of the two trials behind that approval cut the results by age, sex, BMI, starting severity and neck circumference[1]. What the trial measured in the first place is set out in tirzepatide and sleep apnea.

The broad answer is yes. Reductions in the apnea-hypopnea index appeared in every subgroup examined, alongside improvements in body weight, systolic blood pressure and the sleep-apnea-specific hypoxic burden, which measures how much oxygen desaturation someone actually accumulates rather than just counting events.

That matters for anyone converting a trial number into an expectation. A person with an AHI of 18 and a person with an AHI of 60 will both improve on this evidence, and they will not improve by the same amount, and the average reported for the trial describes neither of them. It is the same reason a percentage weight loss from a lighter starting population is not comparable to one from a heavier one, as the ecnoglutide trial illustrates.

The word to hold onto is the authors’ own: descriptive. No interaction tests are reported here and no confidence intervals accompany the subgroup estimates, so “consistent across subgroups” means the ranges overlap rather than that heterogeneity was tested and ruled out. That is a weaker statement than the prespecified sex analysis in the HFpEF trial, where interaction P values were calculated and reported.

None of this touches the question that follows from a good result, which is what happens to the machine beside the bed. Improving on a measurement is not the same as no longer needing treatment, and a sleep study is the only thing that establishes the difference — a practice gap covered in the cardiac imaging evidence from a different direction.

Frequently asked

Does it work equally well for everyone?
Improvements appeared in every subgroup examined, with similar ranges by age, sex, BMI and neck circumference. By starting severity the range was much wider, from 12.1 to 52.2 fewer events per hour.
Why does starting severity matter so much?
Largely arithmetic. Someone with 20 breathing events an hour cannot lose 52 of them, so people who start worse have more room to improve and the trial average reflects how severe the enrolled group was.
Were the subgroups formally compared?
No. The authors describe these analyses as descriptive, and no interaction tests or subgroup confidence intervals are reported — so consistency here means overlapping ranges rather than tested equivalence.

Sources

  1. [1] Falcon B, Xie CC, Redline S, Grunstein R, et al. (2026). Association of tirzepatide with changes in OSA-related measures based on baseline characteristics - post hoc analyses of SURMOUNT-OSA Journal of Clinical Sleep Medicine. PMID 42675225

Where to get it

Best GLP-1 injections

Every injectable seller we can verify, with the price each one publishes and an honest read of what the trials measured.

Compare providers →

More in Evidence