Zepbound for Sleep Apnea: What the Evidence Shows

Zepbound is the first — and so far only — medication approved by the FDA to treat obstructive sleep apnea. The indication was added on 20 December 2024, and the label now reads that Zepbound is indicated, in combination with a reduced-calorie diet and increased physical activity, "to treat moderate to severe obstructive sleep apnea (OSA) in adults with obesity"¹. That is a real, specific, narrow indication, and it is worth reading every word of it: moderate-to-severe, adults, with obesity.

The evidence behind it is strong on the measurement that defines the disease. It is considerably thinner — and much more interesting — on the symptom most people actually come to this topic for, which is being tired all day. This page covers both, in that order, because the gap between them is the part nobody writes about.

What was approved, and on what

Obstructive sleep apnea is graded by the apnea–hypopnea index (AHI), the number of breathing pauses and shallow-breathing events per hour of sleep. It is common, it is heavily underdiagnosed, and excess weight is its dominant modifiable risk factor²³. The approval rests on SURMOUNT-OSA, a pair of 52-week phase 3 randomised, placebo-controlled trials in adults with moderate-to-severe OSA and obesity, published in the New England Journal of Medicine⁴.

One point of housekeeping first, because it trips up almost everyone quoting this literature: the two trials are numbered differently in every source. The NEJM paper calls them trial 1 (participants not on positive airway pressure) and trial 2 (participants on PAP). The Zepbound label calls the same two trials Study 5 and Study 6. The patient-reported-outcomes paper calls them Study 1 and Study 2¹⁴⁵. They are the same two trials throughout. On this page, "off PAP" and "on PAP" are used instead of a number.

The primary endpoint was change in AHI at week 52, and the result is unambiguous. In the off-PAP trial, AHI fell by 25.3 events per hour on tirzepatide versus 5.3 on placebo, a treatment difference of −20.0 events per hour (95% CI −25.8 to −14.2). In the on-PAP trial it fell by 29.3 versus 5.5, a difference of −23.8 (95% CI −29.6 to −17.9)¹⁴. Baseline AHI averaged around 50 events per hour, so these are people starting with severe disease, and the drug cut the index roughly in half.

The composite endpoint the label reports as "remission or mild non-symptomatic OSA" — defined as an AHI below 5, or an AHI of 5 to 14 combined with an Epworth Sleepiness Scale score of 10 or less — was reached by 42.2% versus 15.9% off PAP and 50.2% versus 14.3% on PAP¹. Roughly half of treated patients on PAP no longer met the criteria for symptomatic OSA after a year. By the standards of sleep medicine, where the therapeutic options have been a machine, a mouthpiece, or surgery, that is a genuinely significant result.

The thing the label does not print

Now the part that is missing from every piece of coverage we could find.

The Zepbound label contains no Epworth Sleepiness Scale result. Not a change score, not a treatment difference, not a p-value. Search the label for ESS and you find it exactly twice: once in Table 8, as a baseline characteristic (mean 10.5 and 10.0 in the two trials), and once inside the definition of the composite remission endpoint, where an ESS of 10 or less is one of two conditions that must be met¹. In both places it is an input, never an outcome.

That matters because daytime sleepiness is the symptom that sends people to a sleep clinic in the first place. A patient reading this label learns precisely how much their AHI, hypoxic burden and body weight are expected to change, and learns nothing about whether they will feel less sleepy.

The ESS effect exists — but it lives only in a separate patient-reported-outcomes paper in Sleep Medicine⁵, and its size is sobering:

• •Off PAP: ESS improved by a least-squares mean difference of −1.4 (95% CI −2.5 to −0.3) versus placebo at week 52. Statistically significant.
• •On PAP: −0.9 (95% CI −2.1 to 0.2). The confidence interval crosses zero. Not significant.

And here is the detail that should govern how you read those numbers: that same paper cites a published minimal important difference for the ESS of 2.6⁵. The significant result, −1.4, is roughly half of it. To be fair to the authors, they flag exactly this — they note the 2.6 threshold was derived as half a baseline standard deviation and was intended to classify individual responders rather than to interpret differences between group means, and they caution that MID thresholds vary greatly with method and sample⁵. That caveat is legitimate. It does not change the fact that the only significant sleepiness result in the programme is smaller than the only sleepiness threshold the paper itself puts on the table.

Two further qualifiers from the same paper, both stated by its authors:

ESS was exploratory, not prespecified. Only the PROMIS Sleep-Related Impairment and Sleep Disturbance measures were prespecified secondary endpoints; every other patient-reported outcome, ESS included, was exploratory and "may not have been adequately powered to detect differences between treatment groups"⁵.

The prespecified sleep-impairment results run into the same problem. PROMIS Sleep-Related Impairment improved by −3.4 off PAP and −4.3 on PAP versus placebo — against a cited MID for that instrument of 4.9⁵. Both point estimates fall below it.

There is also a structural reason the sleepiness signal is muted, and the authors name it: only about half the SURMOUNT-OSA population reported excessive daytime sleepiness at baseline despite severe disease by AHI, because participants were recruited from endocrinology and bariatric clinics rather than sleep centres, and no minimum sleepiness level was required to enrol⁵. People who are not sleepy at baseline cannot get less sleepy. That is a floor effect, not a failure of the drug — but it also means these trials were not well positioned to answer the sleepiness question, and the honest reading is that the question remains substantially open.

What this means if you are the patient

The AHI result is solid and the daytime-symptom result is not. If your goal is to reduce apnoeas and the physiological burden of untreated OSA, the evidence supports the drug. If your goal is specifically to stop feeling exhausted and foggy all day, the honest answer is that the trial programme measured that as an afterthought, found an effect below its own cited threshold in one trial and nothing significant in the other, and the label declines to print a number.

It is not a replacement for PAP, and the trials did not test it as one. The on-PAP trial enrolled people already using positive airway pressure, and PAP was withdrawn only for the assessments⁵. Nothing in this programme compared tirzepatide against PAP, or established that a responder can stop PAP. That is a decision for a sleep physician holding a repeat sleep study, not one to make from a label.

The indication requires a real diagnosis. Moderate-to-severe OSA means a polysomnogram or home sleep test, an AHI of at least 15, and adults with obesity. If you have not been tested, the first step is a sleep study, not a prescription — and untreated OSA is worth diagnosing on its own terms, since treating it produces measurable, if modest, cognitive improvement in randomised trials⁶. We cover that link in detail in sleep apnea and brain fog, and if your main complaint is feeling wrecked on waking, why you wake up with brain fog covers the other causes worth ruling in.

A note on why there is no product link on this page: the only medication with this indication is branded Zepbound, obtained through a diagnosed-OSA pathway that runs on insurance prior authorisation. Direct-to-consumer telehealth weight-loss programmes cannot fulfil an OSA prescription, and pointing you at one from an article about a sleep-apnea indication would be a bait and switch. So this page sells nothing. If you meet the criteria, the conversation is with a sleep specialist and a prescriber.

The honest summary

Zepbound is genuinely approved for moderate-to-severe obstructive sleep apnea in adults with obesity, and the AHI evidence behind that approval is strong: reductions of 25 to 29 events per hour against 5 on placebo, with 42% to 50% of treated patients reaching remission or mild non-symptomatic OSA¹⁴. What the approval does not come with is a demonstrated improvement in how sleepy you feel. The label prints no Epworth number at all; the only place that number exists is an exploratory analysis showing −1.4 in one trial against a cited minimal important difference of 2.6, and a non-significant −0.9 in the other, in a population only half of whom were sleepy to begin with⁵. Both facts are true at once, and anyone telling you this drug is proven to fix daytime sleepiness is reading past the evidence — as is anyone dismissing an AHI reduction of that size. If the fatigue itself is what you are chasing, what the FDA labels say about tiredness on a GLP-1 is the companion piece, and what actually causes brain fog covers the drivers worth ruling in alongside a sleep study.