Zepbound for Sleep Apnea: What SURMOUNT-OSA Actually Found

On December 20, 2024 the FDA cleared Zepbound for obstructive sleep apnea — no other drug of any kind carries that indication.^[6] The license is narrow. It covers adults who have obesity as well as OSA at the moderate-to-severe end, and it positions the drug beside a calorie deficit and more activity rather than in place of them. It works: in the two SURMOUNT-OSA trials, breathing interruptions fell by 25.3 and 29.3 an hour against 5.3 and 5.5 on placebo.^[1] What almost no coverage mentions is where that leaves people. Subtract those changes from the trial's own starting points and the average participant still had moderate sleep apnea at week 52. This is a large improvement, and it is not a cure.

What the trials did

SURMOUNT-OSA was run as two separate 52-week trials rather than one. Trial 1 enrolled people who were not using positive airway pressure. Trial 2 enrolled people who were. Each assigned participants to placebo or to tirzepatide, at whichever of the two upper strengths — 10 mg, otherwise 15 mg — the person tolerated. Both measured one primary endpoint: the change in the apnea-hypopnea index, meaning the number of times an hour that breathing stops or goes shallow during sleep.^[1]

Splitting the trial that way was the right call, because the two groups are asking different questions. One is asking what the drug does for untreated apnea. The other is asking what it adds for someone already on a machine.

The published numbers

Tirzepatide also improved every prespecified secondary endpoint against placebo: body weight, hypoxic burden, high-sensitivity C-reactive protein, systolic blood pressure, and patient-reported sleep impairment and disturbance.^[1] Gastrointestinal effects were the most commonly reported adverse events, mostly mild to moderate — the same profile the drug carries at any dose, covered in our constipation guide and the side-effect timeline.

Where that actually leaves you

Here is the step the promotional coverage skips. The trial reports change in AHI, not the number people finished on. Do the subtraction yourself. In trial 1, a mean baseline of 51.5 less a mean change of 25.3 leaves about 26 events an hour. In trial 2, 49.5 less 29.3 leaves about 20.

Sleep apnea is graded as mild at 5 to 14.9 events an hour, moderate at 15 to 29.9, and severe at 30 or more.^[3] So the average participant in both trials moved from severe apnea to moderate apnea. That is a genuinely meaningful shift and it is worth having. It is not the same as no longer having the condition, and someone reading “first drug approved for sleep apnea” could easily believe otherwise.

The average participant went from severe sleep apnea to moderate sleep apnea. Both numbers are still a diagnosis.

There is a related detail worth knowing about trial 2. For those participants, PAP was withdrawn before the polysomnography and the questionnaires were administered.^[2] The trial-2 figures therefore describe what tirzepatide did on its own for people who had been on a machine — not what the drug and the machine achieve together. Nothing in SURMOUNT-OSA measured the combination.

Sleep, as opposed to the index

An apnea-hypopnea index is a count. It is not the same as feeling rested, and a treatment can move one without moving the other. A 2025 analysis of the patient-reported measures collected inside SURMOUNT-OSA looked at exactly this, and found significant improvements against placebo at week 52 in sleep-related impairment, sleep disturbance, daytime functioning and health-related quality of life, across several validated instruments.^[2]

That matters more than it might sound. The count improving and the person feeling no different would be a real possibility, and it is the outcome most drug trials in this space are not designed to detect.

What about the other GLP-1 drugs

Only tirzepatide has the indication. Liraglutide was tested for this a decade ago: SCALE Sleep Apnea randomized 359 adults with obesity and moderate or severe OSA who were unwilling or unable to use PAP, and over 32 weeks the AHI fell by 12.2 events an hour against 6.1 on placebo, with weight down 5.7% against 1.6%.^[3] Real, and a fraction of what tirzepatide produced.

Semaglutide has no dedicated sleep apnea trial. If a seller markets a compounded semaglutide product for sleep apnea, that claim is not standing on a trial in this condition, and marketing an unapproved use is a pattern we track on the FDA warning letters page.

The coverage question

The approval changed who can plausibly get this covered. Before December 2024, a Zepbound prescription was a weight-management prescription, and a great many plans exclude weight-management drugs outright. An OSA diagnosis is a different clinical route, and some plans that refuse the first will consider the second.

It is not automatic. Expect prior authorization, expect a documented sleep study, and expect the plan to have its own criteria for what counts. Our insurance section tracks what individual plans publish, and the prior authorization builder assembles what a submission generally needs. What we cannot tell you is what your specific plan will do, and any seller who says they know is guessing.

For what the drug does on the scale rather than in the sleep lab, our tirzepatide results timeline works through the SURMOUNT weight-loss trials, and the price tracker carries a dated figure for every seller we follow.