What ResMed Quarterly Earnings Actually Say About the Ozempic Threat

In the late summer of 2023, Novo Nordisk released topline results from SELECT, its cardiovascular outcomes trial of semaglutide. ResMed had nothing to do with the study. Its stock fell anyway — roughly a third of its market value over the months that followed, on the theory that a weekly injection would eventually empty the sleep lab.

Several years of ResMed quarterly earnings later, that theory has been tested about as thoroughly as a market thesis ever gets tested in public. The results are stranger than either the bulls or the bears wrote down.

Anchor figures first, and treat them as round: for the fiscal year ended June 30, 2025, ResMed reported revenue near $5.1 billion, up about 10 percent, with a June-quarter print around $1.35 billion at a similar growth rate. Non-GAAP gross margin sat near 60 percent, several hundred basis points above the 2022–23 supply-chain trough, and non-GAAP EPS grew faster than revenue. Pull the current filing before you trade on any of it. The direction is what matters here.

The myth, stated in its strongest form

The bear case was never stupid. It went like this: obstructive sleep apnea is, at the population level, an obesity disease. Roughly 60 to 70 percent of moderate-to-severe cases in Western cohorts occur in people with a BMI above 30. Give those people a drug that reliably removes 15 to 22 percent of body weight, and you shrink the addressable market from the demand side — permanently, and faster than a device company can reprice.

The elegant version added a second blade: CPAP adherence is famously mediocre, hovering somewhere between 50 and 70 percent at one year depending on how generously you define "adherent." A therapy people already dislike, competing against a weekly injection people are willing to pay cash for, is a therapy in structural decline.

That argument survives contact with the biology only partway.

Do GLP-1 drugs cure sleep apnea?

No. They reduce it, substantially, in a subset of patients, and the distinction between reduce and cure is the whole investment case. In the pivotal trial, roughly 40 to 50 percent of participants on tirzepatide reached what the authors called disease resolution or mild residual disease. The rest still had measurable apnea. Many still had enough of it to warrant treatment.

Put differently: the drug moves a large fraction of severe patients into the moderate band and a large fraction of moderate patients into the mild band. It does not empty the band. And a patient who moves from an apnea-hypopnea index of 60 to an AHI of 22 is still a PAP patient in most sleep clinics in the world.

What the trial actually measured

The study is Malhotra et al. (2024), New England Journal of Medicine — two parallel 52-week trials, 469 participants total, adults with obesity and moderate-to-severe OSA, one trial in patients using PAP and one in patients not using it. Tirzepatide at maximum tolerated dose (10 or 15 mg weekly) versus placebo. The primary endpoint was change in AHI.

Mean AHI fell by about 25 events per hour on drug versus roughly 5 on placebo. That is a genuinely large effect, larger than anything bariatric-surgery literature had produced outside of case series. The FDA approved tirzepatide for moderate-to-severe OSA in adults with obesity in December 2024 on the strength of it.

Two caveats belong in your model. First, 469 people over one year tells you almost nothing about durability — GLP-1 weight regain after discontinuation is well documented, and nobody has published five-year apnea data. Second, AHI is a weak surrogate.¹ It counts events without weighting how long they last or how far oxygen falls.

The mechanism, in the order it happens in the airway

Walk it through the way the body does it.

You lie down. Gravity and muscle tone stop protecting the pharynx, a soft tube with no rigid support along most of its length. Fat deposited in the parapharyngeal pads and — this is the underappreciated part — inside the tongue itself narrows that tube from the outside and the inside at once. Wang et al. (2020), American Journal of Respiratory and Critical Care Medicine, imaged 67 patients before and after weight loss and found that reduction in tongue fat was the strongest single mediator of AHI improvement. Not neck circumference. Tongue fat.

Simultaneously, abdominal fat loss raises functional residual capacity. A fuller lung pulls caudally on the trachea, which tensions the pharyngeal wall from below — real physiology, and it explains why apnea improves before much visible weight comes off.

So far, so bearish for device makers. Then the airway closes anyway in a large minority of patients, because anatomy is only one of four endotypes. Eckert et al. (2013), also AJRCCM, phenotyped 75 subjects and found that only about a fifth had anatomy as their sole driver. The other three mechanisms are ones no injection touches:

• Low genioglossus responsiveness — the dilator muscle doesn't recruit hard enough when pressure drops.
• High loop gain — an over-twitchy ventilatory control loop that overshoots CO₂ correction and drives the airway into instability.
• Low arousal threshold — you wake at the slightest respiratory effort, fragmenting sleep before the airway even fully closes.

Weight loss shifts the anatomical term. It leaves the control-system terms roughly where it found them. That is the biological floor under the market, and it is not a marketing claim — it's the reason a lean 55-year-old with an AHI of 30 exists at all.

What the income statement is actually sensitive to

Here the earnings data has been unkind to the simple bear thesis and unkind to the simple bull thesis in different places.

Mask and accessory resupply — the recurring, higher-margin half of the business — has grown steadily rather than eroding. ResMed's management has argued the opposite of the extinction case: that GLP-1 prescriptions drag patients into a physician's office, where undiagnosed apnea gets found. The company has cited its own analysis of de-identified U.S. claims data showing patients prescribed a GLP-1 were meaningfully more likely to start PAP therapy and stay on it, with figures quoted in the neighborhood of ten percentage points. Note the obvious: that analysis is company-funded, retrospective, and confounded by the fact that people who fill expensive prescriptions are people engaged with the medical system. Directionally plausible. Not evidence.

The real headwinds in recent prints have been elsewhere and duller. Philips Respironics, absent from the U.S. sleep market since the June 2021 foam recall and a 2024 consent decree, has been working its way back — a share-donation risk nobody was modeling in 2023. Life-support and ventilation lines have been a drag, including a suspension of new Astral device sales that carried a discrete revenue and provision hit management flagged as one-time. Read those separately from the sleep franchise, because they behave differently.

And the denominator remains enormous. Benjafield et al. (2019), Lancet Respiratory Medicine, estimated 936 million adults aged 30–69 worldwide with mild-to-severe OSA, 425 million of them moderate-to-severe. Treated penetration is in the single digits globally. A drug that removes 40 percent of a diagnosed population is competing against an undiagnosed population an order of magnitude larger.

An honest rule of thumb

Stop watching the headline number for the GLP-1 signal. It's too aggregated to carry the information. Watch mask and resupply revenue against device revenue in the same quarter — if the drug thesis is working, resupply decays before new device sales do, because it reflects the existing installed base leaving.

If resupply is growing and device growth is slowing, you're looking at competition or reimbursement, not pharmacology.

The question nobody has funded

Here is what the data does not tell you, stated plainly. No one has run a trial that randomizes patients with moderate OSA and obesity to PAP alone, GLP-1 alone, or both, and follows them for five years on hard cardiovascular outcomes. Until that exists, the entire debate — the selloff, the recovery, every model built on penetration curves — rests on inference from a surrogate endpoint that sleep researchers have been publicly criticizing since at least 2019.

So the open question isn't whether tirzepatide lowers AHI. It does. It's this: when a patient's AHI falls from 60 to 22 and they feel better, has their heart actually been protected, or have they simply stopped being counted?